Thread: Elocution

So, here is my parable.

   Should you drive to work in an M1A tank? There are lots of very good
reasons to do so, prominantly the way driving to work in an M1A tank
enhances your personal safety. In both freeway incidents and grocery
store parking lots, it is the fellow driving the M1A tank who comes out
on top. However, there are lots of reasons not to drive an M1A tank to
work. The initial aquisition cost of several million dollars is pretty
hard to swallow.  And even if you can quietly steal one from the local
Army base, the fuel costs alone will bankrupt you in short order.

   Companies have been running their IT infrastructures on the
equivalent of M1A tanks for the past several years, and the fuel bill is
starting ot catch up with them.  The first manifestation of this
changeover is the way Linux is eating the bottom out of the proprietary
UNIX market.  Why run your web server on an Ultra 450? It is the finest
hardware around, but it is not actually *needed* for the application.
Between commodity hardware and simple failover systems you can achieve
the same results for far less money. So why not save the money?

   Once you look at how the operating system market is shaking out, the
next chapter seems blindingly obvious.  Oracle is wonderful software,
but it is an M1A tank, and its many features are not *required* for most
applications. Why are people running contact management software on
Oracle? Why are they running web services on Oracle? Like proprietary
UNIX, in many installations Oracle is a nice-to-have, not a
have-to-have. And cost-concious CIOs should be looking with just as much
focus at their Oracle database budgets as they have recently been
looking at their proprietary UNIX budgets.

Has a certain simplicity, doesn't it?

--
       __
      /
      | Paul Ramsey
      | Refractions Research
      | Email: pramsey@refractions.net
      | Phone: (250) 885-0632
      \_

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


> So, here is my parable.
>
>    Should you drive to work in an M1A tank? There are lots of very good

I think that is a faulty analogy. Think of Oracle as a BMW: it's expensive,
but it has a brand name behind it, it is extremely reliable and safe,
and you can take it back to the repair shop when it breaks. And of course,
it looks good to be driving a BMW. ("Powered by Oracle" has a nice
ring to it that adds some authenticity to your site)

Sure, it is expensive and might not be needed, but why should a company
limit its vision by getting a crappy Yugo-SQL that serves its needs now,
but not in two years when the company grows? Oracle may be overkill
for a lot of companies, but it works, it is supported, it is powerful,
and it scales like you wouldn't believe.

PostgreSQL is like a shiny new brand of car that nobody has ever
heard of. Mysterious, cheap, unsupported, rumoured to be powerful.
And it has a funny name to boot. :)

I like the cost-saving angle: while you're dumping your old OS for
Linux, why not dump your old DB for Postgres? Name recognition
and myth-busting needs to happen first and foremost.

Greg Sabino Mullane greg@turnstep.com
PGP Key: 0x14964AC8 200212091538

-----BEGIN PGP SIGNATURE-----
Comment: http://www.turnstep.com/pgp.html

iD8DBQE99QJTvJuQZxSWSsgRArntAJ93A3tOW3tBh2Vtu6ylk1ww659h4QCfRWe0
zog0LIeVhrz+aIDKL5h0OWY=
=A+00
-----END PGP SIGNATURE-----

greg@turnstep.com wrote:

> I like the cost-saving angle: while you're dumping your old OS for
> Linux, why not dump your old DB for Postgres? Name recognition
> and myth-busting needs to happen first and foremost.

    Hm, I started thinking about myth-busting and positioning in general,
and have some less happy thoughts, based on the linux analogy.

    Interestingly, even though Linux is free, the adoption pattern seems
to match Geoffry Moore's "Crossing the Chasm" paradigm very well.
Starting with a beachhead in the web server niche (compelling story: low
cost and high reliability; pitch: it is Solaris without the pricetag;
target market: startup ASPs) and now spreading out into other areas.

    The bad news is, MySQL is already firmly ensconsed in the parallel
niche for web databases, and is starting to spooge outwards from that niche.

    So, if we wanted to pick a niche for PostgreSQL, where would it be?
What is the elevator pitch? In Moore's formulation, how are the blanks
filled in?

"For (target customers -- niche segment)
   Who are dissatisfied with (the current market alternative)
   Our product is a (new product category)
   That provides (key problem solving ability)
   Unlike (the product alternative)
   We have assembled (key whole product features for your
      niche application)"


--
        __
       /
       | Paul Ramsey
       | Refractions Research
       | Email: pramsey@refractions.net
       | Phone: (250) 885-0632
       \_

greg@turnstep.com wrote:

>>   Should you drive to work in an M1A tank? There are lots of very good
>
> I think that is a faulty analogy. Think of Oracle as a BMW: it's expensive,
> but it has a brand name behind it, it is extremely reliable and safe,

Well, that's the analogy Oracle might prefer, but from the perspective
of PostgreSQL soundbites I think the M1A tank works better :) Besides,
if a great nameplate, quality product, and corporate reputation were all
that mattered, Sun wouldn't be looking over their shoulders at Linux so
nervously.

--
       __
      /
      | Paul Ramsey
      | Refractions Research
      | Email: pramsey@refractions.net
      | Phone: (250) 885-0632
      \_

The answer is simple really. Fingers. When something goes wrong fingers turn
into blame compasses. At my two previous jobs and my current one, immunity
of finger pointing was the #1 reason for sticking with something expensive.
This pointing applies in two directions: to managers and to vendors.

If a manager keeps the tanks, and the M1A2 dies in the middle of a
battlefield, they can say, "well we had an acceptable track record with the
M1A1, so it wasn't a crazy idea." Others will follow the reasoning (assuming
the M1A2 has similar design criteria and didn't end up a being a VW beetle)
Reason will prevail, and he will keep his job. If a manager replaces systems
with new ones, he had better justify it on very solid grounds.  If it fails,
he gets canned, the company suffers, fire and brim stone, egg on the face,
etc. Plain and simple. Since the business is used to "buying tanks" and
tanks have always worked for them so far, staying with tanks is a safe move.
They know the fuel consumption and repair rates. The staff is trained on
tanks. Switching to Hyundias is very risky. (Sorry for the cheap shot, but
more on this later.)

When things go wrong in an immediate nature, the manager can call the vendor
and get support. While this is largely true with most important open-source
products today, it seems not as good. The people who support don't always
own the code, and there is no guarantee that the fix/hack will make it into
the source tree, no matter how benevolent the dictator. Maybe new patches
need to be developed and applied for the next version. Also, trying to
support things yourself is costly, and usually incomplete. (Though as a
hacker, I prefer doing it myself!)

Back to the Hyundai remark. OSS is inexpensive, light, stable and while
not-new, the image has been only recently grown to the point that people are
very familiar with the product line. (Elantra, Tiburon, Santa Fe)->(Linux,
Apache, MySQL*) (*by popularity)

Lastly, when everyone else is driving M1A1's, would you feel safe on the
information super-highway even though you saved $989,999?  I'd be wondering
about the situation I put myself in, and what I was smoking when I did.
Yeah, I've got a lot more money than everyone else to spend now, but it
doesn't do you any good if you're a pancake on the information
super-highway.

Fortunately, all those reasons are merely psychological - not technical -
and attitudes will/can change. It's not even MS FUD too. When I told my
current boss about PostgreSQL, and it being free, he asked: "What's wrong
with it? How can it be free?" To which I answered: "Well you can pay for it
if you want." It didn't have the same effect. (Note to self: Set up an S
corp., charge for PostgreSQL disks, manuals and 30-day installation support,
$10000 (Starbucks effect))  Eventually people will see the Hyundias running
circles around the tanks, and people will feel a bit safer about buying one.





-----Original Message-----
From: pgsql-advocacy-owner@postgresql.org
[mailto:pgsql-advocacy-owner@postgresql.org]On Behalf Of Paul Ramsey
Sent: Monday, December 09, 2002 2:39 PM
To: pgsql-advocacy@postgresql.org
Subject: [pgsql-advocacy] Elocution



So, here is my parable.

   Should you drive to work in an M1A tank? There are lots of very good
reasons to do so, prominantly the way driving to work in an M1A tank
enhances your personal safety. In both freeway incidents and grocery
store parking lots, it is the fellow driving the M1A tank who comes out
on top. However, there are lots of reasons not to drive an M1A tank to
work. The initial aquisition cost of several million dollars is pretty
hard to swallow.  And even if you can quietly steal one from the local
Army base, the fuel costs alone will bankrupt you in short order.

   Companies have been running their IT infrastructures on the
equivalent of M1A tanks for the past several years, and the fuel bill is
starting ot catch up with them.  The first manifestation of this
changeover is the way Linux is eating the bottom out of the proprietary
UNIX market.  Why run your web server on an Ultra 450? It is the finest
hardware around, but it is not actually *needed* for the application.
Between commodity hardware and simple failover systems you can achieve
the same results for far less money. So why not save the money?

   Once you look at how the operating system market is shaking out, the
next chapter seems blindingly obvious.  Oracle is wonderful software,
but it is an M1A tank, and its many features are not *required* for most
applications. Why are people running contact management software on
Oracle? Why are they running web services on Oracle? Like proprietary
UNIX, in many installations Oracle is a nice-to-have, not a
have-to-have. And cost-concious CIOs should be looking with just as much
focus at their Oracle database budgets as they have recently been
looking at their proprietary UNIX budgets.

Has a certain simplicity, doesn't it?

--
       __
      /
      | Paul Ramsey
      | Refractions Research
      | Email: pramsey@refractions.net
      | Phone: (250) 885-0632
      \_


---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
    (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)

Here's the thing: I agree 100% with everything you say, all of it. And
yet, right now in the area of operating systems, corporations are
retiring their tanks (proprietary UNIX) in favour of Hyundi (Linux).
What gives? Something has changed, and it is something we should look
very closely at.

My guess as to what as changed is that some of the tank-makers have
abandoned the field and are now marketing the Hyundi as a Humvee (to
further complicate the analogy).  IBM primarily, but also Dell, Compaq
and even Sun (!!?!?! Better Red than Dead, perhaps?) have given Linux
the stamp of corporate "acceptability" and that has been what put it
over the top.

So, who is going to be our IBM? I thought Red Hat might be it, but they
are too dependant on Oracle right now to risk offending them. IBM itself
already has two databases, it is hard to imagine them taking on another
one (of course, they already had a couple operating systems before
taking on Linux). HP/Compaq is a possibility. But really, it will take a
company with existing credibility and accounts to really push things
over the top.

Here's the bad news again though: In the Linux/Apache case, IBM took the
OSS product with the most groundswell momentum and surfed the wave. They
did not make the wave themselves, they just amplified it. In the db
sphere, the groundswell product is not pgsql unfortunately.

That's my thesis anyways. Thoughts?


Jason Hihn wrote:
> The answer is simple really. Fingers. When something goes wrong fingers turn
> into blame compasses. At my two previous jobs and my current one, immunity
> of finger pointing was the #1 reason for sticking with something expensive.
> This pointing applies in two directions: to managers and to vendors.
>
> If a manager keeps the tanks, and the M1A2 dies in the middle of a
> battlefield, they can say, "well we had an acceptable track record with the
> M1A1, so it wasn't a crazy idea." Others will follow the reasoning (assuming
> the M1A2 has similar design criteria and didn't end up a being a VW beetle)
> Reason will prevail, and he will keep his job.

--
       __
      /
      | Paul Ramsey
      | Refractions Research
      | Email: pramsey@refractions.net
      | Phone: (250) 885-0632
      \_

Paul Ramsey wrote:
> greg@turnstep.com wrote:
<snip>
>    So, if we wanted to pick a niche for PostgreSQL, where would it be?
> What is the elevator pitch? In Moore's formulation, how are the blanks
> filled in?

When people have been asking what the difference is between MySQL and
PostgreSQL, thus far I have been explaining the differences between a
web-specific database and an "application level" database, then drawing
parallels between PostgreSQL and the other well known DB's (Oracle, DB2,
Sybase, etc).

It's worked well for the people that I've been explaining this too thus
far, as they seem to grasp the differences pretty easily, etc.

Would this be an angle we could strongly push?

:-)

> "For (target customers -- niche segment)

Application level database market segment?


>   Who are dissatisfied with (the current market alternative)

The high license cost, high staff cost, etc of the leading
proprietary/commercial-only alternatives.

Maybe a strong branding point would be to call the proprietary databases
"commercial-only", as it does have a mental association with
"inflexibility" built in.  Might be able to leverage that significantly
in some way?


>   Our product is a (new product category)

An "application level database".  Don't think it's a new product
category, although we are in a class of software that some people in the
traditional software markets are unfamiliar with.


>   That provides (key problem solving ability)

As per traditional application level database stuff.  :)


>   Unlike (the product alternative)

Database products built for other target markets, with feature sets not
suitable to application level products.


>   We have assembled (key whole product features for your
>      niche application)"

Hmm.... seems like a worthwhile idea/approach.


Regards and best wishes,

Justin Clift



--
"My grandfather once told me that there are two kinds of people: those
who work and those who take the credit. He told me to try to be in the
first group; there was less competition there."
- Indira Gandhi

Justin Clift wrote:
>
> When people have been asking what the difference is between MySQL and
> PostgreSQL, thus far I have been explaining the differences between a
> web-specific database and an "application level" database, then drawing
> parallels between PostgreSQL and the other well known DB's (Oracle, DB2,
> Sybase, etc).

Is "application level" database commonly understood terminology? Would
saying "MySQL is a fine web database, but for many applications you need
an enterprise database like PostgreSQL" be too far beyond the pale?

> Would this be an angle we could strongly push?

"PostgreSQL - The open source database which really is."

It's pejorative for sure. The question is who are competitors really
are. It is tempting to take on Oracle, because that is where the big
market segment is, at the bottom end of the enterprise. However, perhaps
the most important early battle is against our OSS compatriots, in order
to become the OSS database-of-choice first.

"For cost-conscious data managers,
  Who are dissatisfied with proprietary pricing and licence restrictions,
  Our product is a enterprise-ready open source database,
  That provides all the functionality of proprietary products,
  Unlike MySQL, which is best used in limited web-based roles."


It's really hard to do, particularly when you're not getting $15000 per
licence and rolling 25% of that back into marketing :) The SEC filings
from Oracle were really revealing about just how much money goes into
marketing and how much into R&D. Pull up the Oracle annual report some
time and have a look. Your licencing dollars at work. :)

--
       __
      /
      | Paul Ramsey
      | Refractions Research
      | Email: pramsey@refractions.net
      | Phone: (250) 885-0632
      \_

>     The bad news is, MySQL is already firmly ensconsed in the parallel
> niche for web databases, and is starting to spooge outwards from that
niche.

I'm quite active on a number of database help sites and I think that
Postgres is gaining recognition, as well as MySQL.  There are more articles
that mention Postgres and more people moving to Postgres.

We should think of MySQL as expanding our market.

They are the ones getting the newbies in at the bottom end.  People who
before would never have been brave enough to touch a DBMS are now getting
their hands dirty.  MySQL is making databases accessible.  This is only good
for Postgres.  Just think small piece, but bigger pie!

As these newbies gain in ability, they start to push the database hard and
then they come up against a brick wall when they realise that MySQL simply
cannot do what they want it to do.  Then they turn to PostgreSQL.  I've seen
it time and time again in the newbie database forums.

Someone who has been using MySQL for years on a personal basis would be
happy, if it was required, to support a PostgreSQL installation.  This is
because they know about open source databases and they know that they could
have a good crack at it.  This is good for PostgreSQL because a junior admin
with no DBMS experience at all is likely to be scared of doing that and will
instead push for a commercial (Oracle, MSSQL) solution where they have no
responsibility.

That's my thoughts.

Chris

On Mon, 9 Dec 2002, Christopher Kings-Lynne wrote:

> >     The bad news is, MySQL is already firmly ensconsed in the parallel
> > niche for web databases, and is starting to spooge outwards from that
> niche.
>
> I'm quite active on a number of database help sites and I think that
> Postgres is gaining recognition, as well as MySQL.  There are more articles
> that mention Postgres and more people moving to Postgres.
>
> We should think of MySQL as expanding our market.

Right. Its always better to be the underdog, as it were. Besides,
Sleepycat DB is *by far* the most successful open source database (in
fact, it is the most widely deployed database).

Once you start comparing PostgreSQL and MySQL you miss the point. We want
to push PostgreSQL for what its good at. We will find far more success
that way.

Gavin

Gavin Sherry wrote:
> On Mon, 9 Dec 2002, Christopher Kings-Lynne wrote:
<snip>
>>We should think of MySQL as expanding our market.
>

We might want to be wary of counting on MySQL continuing to funnel
people our way into the future.  Ever since they started to become
"financially successful" (from what I can tell) they've been far less
inclined to play fair in the Open Source Community.  Can't count how
many times I've asked them privately and publically to update their
pages on the main MySQL site in relation to PostgreSQL.  Think the only
way *that* is going to happen is with legal assistance.  :-(

As they continue to receive funding, they would have to be introducing
ways of adding features (even if badly integrated) to stop people having
to look elsewhere.

:-/

> Right. Its always better to be the underdog, as it were. Besides,
> Sleepycat DB is *by far* the most successful open source database (in
> fact, it is the most widely deployed database).
>
> Once you start comparing PostgreSQL and MySQL you miss the point. We want
> to push PostgreSQL for what its good at. We will find far more success
> that way.

Yep.  Let's aim to gain acceptance from the enterprise (and
governmental) database markets as an effective replacement for many of
their present and future deployments.

:-)

Regards and best wishes,

Justin Clift


> Gavin


--
"My grandfather once told me that there are two kinds of people: those
who work and those who take the credit. He told me to try to be in the
first group; there was less competition there."
- Indira Gandhi

On Tue, 10 Dec 2002, Justin Clift wrote:

> Gavin Sherry wrote:
> > On Mon, 9 Dec 2002, Christopher Kings-Lynne wrote:
> <snip>
> >>We should think of MySQL as expanding our market.
> >
>
> We might want to be wary of counting on MySQL continuing to funnel
> people our way into the future.  Ever since they started to become
> "financially successful" (from what I can tell) they've been far less
> inclined to play fair in the Open Source Community.  Can't count how

That is not true IMHO. As I detailed in another email, I meet up with
MySQL AB people regularly -- Zak Graent, David Axmark, Georg Richter. They
are great guys. Zak in particular does a lot for open source and open
source databases including organising of database forums at OS conferences
and work in the community to enable open source databases work better
together (see my email on advocacy).

> many times I've asked them privately and publically to update their
> pages on the main MySQL site in relation to PostgreSQL.  Think the only
> way *that* is going to happen is with legal assistance.  :-(

That is way too big a call. Justin: stop worrying about this kind of
thing! Compare their Web site, which *merely* compares MySQL to an old
version of Postgres, to the advocacy Web site which makes *many* dubious
claims about Postgres (see my email on advocacy).

> As they continue to receive funding, they would have to be introducing
> ways of adding features (even if badly integrated) to stop people having
> to look elsewhere.

Stupidest thing I've heard today. Are you seriously suggesting that
MySQL's addition of features based on user requirement is any different
than what happens over here?

I'm all for enthusiasm, but not this kind.

Gavin

Gavin Sherry wrote:
> On Tue, 10 Dec 2002, Justin Clift wrote:
>
>
>>Gavin Sherry wrote:
>>
>>>On Mon, 9 Dec 2002, Christopher Kings-Lynne wrote:
>>
>><snip>
>>
>>>>We should think of MySQL as expanding our market.
>>>
>>We might want to be wary of counting on MySQL continuing to funnel
>>people our way into the future.  Ever since they started to become
>>"financially successful" (from what I can tell) they've been far less
>>inclined to play fair in the Open Source Community.  Can't count how
>
>
> That is not true IMHO. As I detailed in another email, I meet up with
> MySQL AB people regularly -- Zak Graent, David Axmark, Georg Richter. They
> are great guys. Zak in particular does a lot for open source and open
> source databases including organising of database forums at OS conferences
> and work in the community to enable open source databases work better
> together (see my email on advocacy).

Please ask them to contact us, as I've directly asked Monty a *bunch* of
times to update stuff, both privately and publically.



>>many times I've asked them privately and publically to update their
>>pages on the main MySQL site in relation to PostgreSQL.  Think the only
>>way *that* is going to happen is with legal assistance.  :-(
>
>
> That is way too big a call. Justin: stop worrying about this kind of
> thing! Compare their Web site, which *merely* compares MySQL to an old
> version of Postgres, to the advocacy Web site which makes *many* dubious
> claims about Postgres (see my email on advocacy).

Hmmm.... seriously don't agree with you here.


>>As they continue to receive funding, they would have to be introducing
>>ways of adding features (even if badly integrated) to stop people having
>>to look elsewhere.
>
>
> Stupidest thing I've heard today. Are you seriously suggesting that
> MySQL's addition of features based on user requirement is any different
> than what happens over here?


Absolutely not.  That's why I said "they would have to be".  It's no
different from anywhere else, as it's part of the natural growth of a
project.


> I'm all for enthusiasm, but not this kind.

Please remove your cynicism hat.

;-/

Regards and best wishes,

Justin Clift


> Gavin
>


--
"My grandfather once told me that there are two kinds of people: those
who work and those who take the credit. He told me to try to be in the
first group; there was less competition there."
- Indira Gandhi

On Tue, 10 Dec 2002, Justin Clift wrote:

> > That is not true IMHO. As I detailed in another email, I meet up with
> > MySQL AB people regularly -- Zak Graent, David Axmark, Georg Richter. They
> > are great guys. Zak in particular does a lot for open source and open
> > source databases including organising of database forums at OS conferences
> > and work in the community to enable open source databases work better
> > together (see my email on advocacy).
>
> Please ask them to contact us, as I've directly asked Monty a *bunch* of
> times to update stuff, both privately and publically.

No. I've just re-read the part of the manual you mentioned. Though I don't
agree with it, I also don't agree with Oracle PR's comments that
'databases are too sophisticated for people working in open source to
successfully develop'. What's our response? We do what we do and *show*
that Postgres is good. We need to do the same thing here. If MySQL AB
contacted you, asking you to remove the 'most advanced opensource database
system in the world' slogan which is all over PostgreSQL Web sites, would
you even consider it? After all, it is simply *our* opinion that this is
the case. I can see convincing arguments which would suggest that MySQL,
SAP/DB, Firebird or Sleepycat DB are actually the most sophisticated. It
is important to understand that.

Gavin

On Tue, 10 Dec 2002, Justin Clift wrote:

> inclined to play fair in the Open Source Community.  Can't count how
> many times I've asked them privately and publically to update their
> pages on the main MySQL site in relation to PostgreSQL.

Isn't the benchmark itself available so that one person coulp spend a
weekend comparing pg 7.3 with whatever is the latest mysql at the moment.
Then there is of course the problem that the benchmark is made to show the
good side of mysql, but I would be interested in the result just because
of that.

--
/Dennis

Hi Dennis,

Dennis Björklund wrote:
> On Tue, 10 Dec 2002, Justin Clift wrote:
>
>
>>inclined to play fair in the Open Source Community.  Can't count how
>>many times I've asked them privately and publically to update their
>>pages on the main MySQL site in relation to PostgreSQL.
>
>
> Isn't the benchmark itself available so that one person coulp spend a
> weekend comparing pg 7.3 with whatever is the latest mysql at the moment.

Would you have the time and inclination to do this?



> Then there is of course the problem that the benchmark is made to show the
> good side of mysql, but I would be interested in the result just because
> of that.

Agreed.  Monty mentioned in about July 2001 that they'd hired someone
(Anna something from memory) to make their benchmark test in a
multi-user fashion instead of a single user process.  Hopefully that's
running nicely now.  The OSDB software (another Open Source database
testing tool) has added the ability to generate large datasets in recent
CVS, so that's another avenue for testing with as well.

:-)

Regards and best wishes,

Justin Clift

--
"My grandfather once told me that there are two kinds of people: those
who work and those who take the credit. He told me to try to be in the
first group; there was less competition there."
- Indira Gandhi

Justin Clift wrote:
> The OSDB software (another Open Source database testing tool) has added the
> ability to generate large datasets in recent CVS, so that's another avenue
> for testing with as well.

Just one note on this. If we ever get serious about using OSDB to compare
Postgres with MySQL, it needs some hacking in order to get a fair comparison.

The problem as I recall is that the two longest running tests (and therefore
the ones with the largest impact on the overall result) are designed to fetch
rows one-at-a-time. The Postgres version does this by using a CURSOR and FETCH
1, while the MySQL version has to fake it using the MySQL client library to
buffer the rows. In other words, the MySQL version of this benchmark is uses a
local library and cache, instead of making a round trip to the server for each
row (something like 10000 times)!

To make the comparison fair, the Postgres version should FETCH the entire
result once using libpq, and then dole out the rows one-at-a-time similar to
the MySQL version.

Joe

Joe Conway wrote:
> Justin Clift wrote:
>
>> The OSDB software (another Open Source database testing tool) has
>> added the
>> ability to generate large datasets in recent CVS, so that's another
>> avenue
>> for testing with as well.
>
>
> Just one note on this. If we ever get serious about using OSDB to
> compare Postgres with MySQL, it needs some hacking in order to get a
> fair comparison.
>
> The problem as I recall is that the two longest running tests (and
> therefore the ones with the largest impact on the overall result) are
> designed to fetch rows one-at-a-time. The Postgres version does this by
> using a CURSOR and FETCH 1, while the MySQL version has to fake it using
> the MySQL client library to buffer the rows. In other words, the MySQL
> version of this benchmark is uses a local library and cache, instead of
> making a round trip to the server for each row (something like 10000
> times)!
>
> To make the comparison fair, the Postgres version should FETCH the
> entire result once using libpq, and then dole out the rows one-at-a-time
> similar to the MySQL version.

Ok, that makes sense.  Andy (the author of OSDB) seems to be happy to
accept suggestions like this.  He has knocked back suggested
improvements (and patches) that he feels change the way OSDB runs in
relation to the AS3AP spec, but this sounds like it's would be far more
efficient and also not change the principle of things.

Would it be cool to forward this suggestion to Andy?

:-)

Regards and best wishes,

Justin Clift


> Joe
>
>


--
"My grandfather once told me that there are two kinds of people: those
who work and those who take the credit. He told me to try to be in the
first group; there was less competition there."
- Indira Gandhi

Justin Clift wrote:
> Ok, that makes sense.  Andy (the author of OSDB) seems to be happy to
> accept suggestions like this.  He has knocked back suggested
> improvements (and patches) that he feels change the way OSDB runs in
> relation to the AS3AP spec, but this sounds like it's would be far more
> efficient and also not change the principle of things.
>
> Would it be cool to forward this suggestion to Andy?
>

Sure -- does it make enough sense that he'll be able to figure out what I'm
trying to say? I guess if he needs more info, have him contact me directly. It
may have even been fixed by now -- been a while since I looked.

Joe

On Tue, 2002-12-10 at 01:02, Paul Ramsey wrote:
> and even Sun (!!?!?! Better Red than Dead, perhaps?) have given Linux

As a guess, I would say that they view themselves as a hardware company
(which will be their death, of course, as their hardware is overpriced,
but I disgress).

cheers
-- vbi

--
this email is protected by a digital signature:  http://fortytwo.ch/gpg

NOTE: keyserver bugs! get my key here: https://fortytwo.ch/gpg/92082481

On Monday 09 Dec 2002 8:44 pm, greg@turnstep.com wrote:
> > So, here is my parable.
> >
> >    Should you drive to work in an M1A tank? There are lots of very good
>
> I think that is a faulty analogy. Think of Oracle as a BMW: it's expensive,
> but it has a brand name behind it, it is extremely reliable and safe,
> and you can take it back to the repair shop when it breaks. And of course,
> it looks good to be driving a BMW. ("Powered by Oracle" has a nice
> ring to it that adds some authenticity to your site)
>
> Sure, it is expensive and might not be needed, but why should a company
> limit its vision by getting a crappy Yugo-SQL that serves its needs now,
> but not in two years when the company grows?

A much better comparison (from our angle) is that existing DB suppliers are
like the British and US car manufacturers of the 70's. It's what the public
have always had so they're happy to pay for it.

We are the Japanese - quietly plugging away in our niche, spending our efforts
on improving the product rather than draping bikin-clad blondes over the
bonnet. A few independently minded souls have done a price/performance
comparison and decided it's what they need. Eventually, word of mouth spreads
and suddenly the outsider becomes a serious contender.

The existing suppliers don't vanish, but they have to raise their game to
compete.

--
  Richard Huxton

> On Mo-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


> A much better comparison (from our angle) is that existing DB suppliers
> are like the British and US car manufacturers of the 70's. It's what the
> public have always had so they're happy to pay for it.
>
> We are the Japanese - quietly plugging away in our niche, spending our
> efforts on improving the product rather than draping bikin-clad blondes

Not to beat this analogy to death, but the implication that we are somehow
"improving" on Oracle is a very dangerous one. I've used Oracle quite a bit,
and they have a product that is still way ahead of everyone else's, even if
they did not have marketing and advertising budget that they do. As wonderful
as Postgres is, we are going to have to go after Oracle on price, and go
after everyone else on features. Discounting anyone as a competetitor is a
foolhardy assumption. I think only Oracle and DB2 are better than Postgres
as far as features, power, and reliability, so we need to continue to strive
for their level and stress the open source and free software angles. MySQL
is the only real free (as in beer) competitor, so we need to stress the fact
that Postgres is a much better database. Others (e.g. MSSQL and Sybase)
should be easier to handle - Postgres is cheaper and a better product.
More to say on MySQL, but I will leave that for another thread.

Greg Sabino Mullane greg@turnstep.com
PGP Key: 0x14964AC8 200212101024

-----BEGIN PGP SIGNATURE-----
Comment: http://www.turnstep.com/pgp.html

iD8DBQE99gdYvJuQZxSWSsgRAkRNAJ9Z9WOZ6Sr+dzUZs3VF9len+aGSggCfTV/Y
RCQhkSE30Cl/z8sGfn9ZXvQ=
=v42J
-----END PGP SIGNATURE-----

Don't forget Red Hat Database is PostgreSQL! I think that if Postgres would
get the same press that (undeservingly) MySQL gets, we'd be #1. And would be
used in the same sentence as Oracle and DB2.

I think we still could have IBM hawk Postgres to some degree. Maybe for a
light version of WebSphere. This would take some convincing to get them to
do, because it defers money to them, and it's likely that they'll never need
DB2. On the other hand WebShere is no easy mess to get out of.

There's an announcement at /. about Mono shipping ASP.Net. The list that
Izcaza gave for supported databases is: "Oracle, MS SQL, Sybase, ODBC,
OleDB, Gnome Data Access, SqLite, MySQL and of course, Postgres." Now I'm
all for saving the best for last, but I'd prefer Postgres to have been
first.

How do we accomplish that? We need to infiltrate high-visibility projects
and be vocal. MySQL is up there because people talked about it, and now a
lot of people use it.

Does anyone know if we can do a "mysql_dump | psql" or similar for migration
(at least for data)? (Are the syntax quirks compatible enough to make it
effortless?)

I feel that we must examine the reasons why MySQL got picked up and ran
with. I don't know other people's reasons, but here are mine:
1) I was learning PHP and MySQL was commonly talked about, and
well-supported
2) MySQL was fast, light and easy to set up. This was when Postgres had a
TOC for INSTALL (6.x).
3) While learning, I knew nothing about databases. Referential integrity
constraints?, triggers? Had I known what they are and how [important it is]
to use them I'd have not used MySQL.
4) MySQL ran under windows almost as easily as under Linux, as did Apache
and PHP. Now you can do it with PostgreSQL, but still, currently, it is no
where as easy.
5) Can't drop a column in Postgres (easily)
6) (though not a concern of mine at the time, it eventually became one)
Replication (MySQL has a binary stream that can be fed to another machine)

While much of the above is now dated, the people who looked into Postgres
probably have bad memories, and a descent investment in MySQL. New awareness
and easy migration are key to getting them back. I came back because I
learned how important ref. integ. constr. were, and how awesome triggers can
be.

Hope that helps
-J

-----Original Message-----
From: Paul Ramsey [mailto:pramsey@refractions.net]
Sent: Monday, December 09, 2002 7:03 PM
To: Jason Hihn; pgsql-advocacy@postgresql.org
Subject: Re: [pgsql-advocacy] Elocution


Here's the thing: I agree 100% with everything you say, all of it. And
yet, right now in the area of operating systems, corporations are
retiring their tanks (proprietary UNIX) in favour of Hyundi (Linux).
What gives? Something has changed, and it is something we should look
very closely at.

My guess as to what as changed is that some of the tank-makers have
abandoned the field and are now marketing the Hyundi as a Humvee (to
further complicate the analogy).  IBM primarily, but also Dell, Compaq
and even Sun (!!?!?! Better Red than Dead, perhaps?) have given Linux
the stamp of corporate "acceptability" and that has been what put it
over the top.

So, who is going to be our IBM? I thought Red Hat might be it, but they
are too dependant on Oracle right now to risk offending them. IBM itself
already has two databases, it is hard to imagine them taking on another
one (of course, they already had a couple operating systems before
taking on Linux). HP/Compaq is a possibility. But really, it will take a
company with existing credibility and accounts to really push things
over the top.

Here's the bad news again though: In the Linux/Apache case, IBM took the
OSS product with the most groundswell momentum and surfed the wave. They
did not make the wave themselves, they just amplified it. In the db
sphere, the groundswell product is not pgsql unfortunately.

That's my thesis anyways. Thoughts?


Jason Hihn wrote:
> The answer is simple really. Fingers. When something goes wrong fingers
turn
> into blame compasses. At my two previous jobs and my current one, immunity
> of finger pointing was the #1 reason for sticking with something
expensive.
> This pointing applies in two directions: to managers and to vendors.
>
> If a manager keeps the tanks, and the M1A2 dies in the middle of a
> battlefield, they can say, "well we had an acceptable track record with
the
> M1A1, so it wasn't a crazy idea." Others will follow the reasoning
(assuming
> the M1A2 has similar design criteria and didn't end up a being a VW
beetle)
> Reason will prevail, and he will keep his job.

--
       __
      /
      | Paul Ramsey
      | Refractions Research
      | Email: pramsey@refractions.net
      | Phone: (250) 885-0632
      \_

On Tue, 2002-12-10 at 10:42, Jason Hihn wrote:
> I feel that we must examine the reasons why MySQL got picked up and ran
> with.

Several of the core group of php developers are windows guys. Since
mysql ran on windows, they made sure php had extremely tight mysql
integration. mysql rode the coat-tails of php to mass popularity.

Robert Treat

I think that is a very astute observation.  IMHO the only advantage
mysql has left is the very most important one, windows compatibility.  I
think many people underestimate just how distasteful it is for most
windows users it is to run a cygwin installation of anything, much less
a production database (even for development purposes).  If you hop over
to the ports and cygwin lists you will find tons of issues, even very
basic ones, with cygwin installations.

I personally think a win32 will instantly catapult postgresql into
superstar status and bring in legions of developers who are dissatisfied
with Microsoft products.  It will also make postgres a member of a very
small club of full feature databases that you can package and ship with
a commercial application.

Merlin




-----Original Message-----
From: Robert Treat [mailto:xzilla@users.sourceforge.net]
Sent: Tuesday, December 10, 2002 1:25 PM
To: Jason Hihn
Cc: pgsql-advocacy@postgresql.org
Subject: Re: [pgsql-advocacy] Elocution

On Tue, 2002-12-10 at 10:42, Jason Hihn wrote:
> I feel that we must examine the reasons why MySQL got picked up and
ran
> with.

Several of the core group of php developers are windows guys. Since
mysql ran on windows, they made sure php had extremely tight mysql
integration. mysql rode the coat-tails of php to mass popularity.

Robert Treat



---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/users-lounge/docs/faq.html

> I personally think a win32 will instantly catapult postgresql into
> superstar status and bring in legions of developers who are dissatisfied
> with Microsoft products.  It will also make postgres a member of a very
> small club of full feature databases that you can package and ship with
> a commercial application.

I can think of no other Great Database for Windows. MySQL might come close,
but we can sweep them under the rug by pointing out that moving between
(develop, deploy) on PG is much easier because of similar feature sets. How
in MySQL can you do triggers and ref. integ. constraints? Any seasoned app
developer will know that each customer has a preference for a particular
database. Working from one that is highly ANSI compliant, as well as
possessing the same features as the Great Databases, will ease the portings.
This is a completely reasonable FUD tactic. You never know what DB your
customer is going to require it in. One may want Oracle, another SQL Server,
another DB2. The cost of developing against all of those (licensing only) is
roughly $50,000, where as PostgreSQL is free, freely deployable and easily
ported from.  (What my former company did when they found themselves in this
situation is to get a commitment that would cover the cost. Alternatives are
to have the customer ship you the machine with the DB pre-installed for
testing if support/maintenance is not required.)

Even though you could use MySQL in the same way, (putting trigger logic in
the app) one problem is that if you have multiple clients, all clients must
be upgraded lest one put your data asunder. By developing business rules and
following them, you only need to update the clients when the business logic
changes. Other times that enhancements are made, they are made in 1 place,
and all others get the updates intrinsically, instantly. That is another FUD
item.

Note: FUD is ok, as long as it's valid.

But anyway, I agree with Merlin, particularly that Mono ASP.Net is out
now...


-----Original Message-----
From: Robert Treat [mailto:xzilla@users.sourceforge.net]
Sent: Tuesday, December 10, 2002 1:25 PM
To: Jason Hihn
Cc: pgsql-advocacy@postgresql.org
Subject: Re: [pgsql-advocacy] Elocution

On Tue, 2002-12-10 at 10:42, Jason Hihn wrote:
> I feel that we must examine the reasons why MySQL got picked up and
ran
> with.

Several of the core group of php developers are windows guys. Since
mysql ran on windows, they made sure php had extremely tight mysql
integration. mysql rode the coat-tails of php to mass popularity.

Robert Treat



---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/users-lounge/docs/faq.html

---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
    (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)

Changing the topic here, you touched on another point that I think could
use some refinement.  I've always thought that not enough emphasis gets
placed on stored procedures and postgres's approach to them.  I hear a
lot of talk about ref. integrity, business rules, and the like (and ansi
compliance, personally, I could give a hoot about that).  SP's and their
ability to be run both inside and outside of a query are whats going to
define postgres's rule in a increasingly .net dominated database world
(in win32).

Ado.net plays the lowest common denominator of database feature sets and
moves all real work to the middleware level.  SPs are a way of stealing
a little bit of that thunder back, and giving the database a little
glory.  (Ado.net also has a very elegant for writing high performance
drivers for 3rd party databases, so is a natural fit for pg).

Case in point: I worked on a GIS application and loved how I could
integrate our library of C and assembler code (like great circle
distance algorithm) into sp's so these functions could be utilized
THROUGH the query at the set based level, instead of lame, slow, error
prone filtering at the middleware level.  Its no accident PostGIS is not
called 'MyGIS' :)  This means my datasets are properly calculated and
filtered before getting bogged down by xml conversions and other cruft
like that.

According to me, SPs and server-side programming interfaces are why
postgres will be the open source database of choice.  (even more
important is the liberal licensing)
Ref. Int. is nice, but you are competing with middleware technologies in
that respect.
As for ANSI compliance, well let's just say I've never read an ansi
standard, and don't plan to :).  (It really is very important, I'm just
being trite).

Note: this also goes for views, but the last time I checked (7.1.3),
views did not optimize the where condition on the view through the
static query, which made them impractical for large datasets.  Still,
something is better than nothing.

Merlin






-----Original Message-----
From: Jason Hihn [mailto:jhihn@paytimepayroll.com]
Sent: Tuesday, December 10, 2002 2:12 PM
To: pgsql-advocacy@postgresql.org
Subject: Re: [pgsql-advocacy] Elocution

>Even though you could use MySQL in the same way, (putting trigger logic
in
>the app) one problem is that if you have multiple clients, all clients
must
>be upgraded lest one put your data asunder. By developing business
rules >and
>following them, you only need to update the clients when the business
logic
>changes. Other times that enhancements are made, they are made in 1
place,
>and all others get the updates intrinsically, instantly. That is
another >FUD
>item.