Thread: Re: PostgreSql and VMS operating System
Bruce Momjian <bruce@momjian.us> writes: > On Fri, Jan 24, 2025 at 01:07:56AM -0800, Christophe Pettus wrote: >> So, basically, if you want a maintained VMS port, you need to either drive the project yourself, or find others who will. > This email thread from 2003 says VMS probably doesn't work anymore > because of lack of testers: > https://www.postgresql.org/message-id/flat/200301071531.h07FVWI08147%40candle.pha.pa.us#0dbc1439f51ec7842125fb8ae200b6da I doubt we ever had a working VMS port. There are precisely zero references to VMS in our commit log, so certainly there was never one that got removed. It's barely possible that PG "just worked" without any patches under their POSIX emulation layer, but I could not find any indication of successful users of PG-on-VMS in the mail list archives either. What I did find was occasional suggestions that we port to OpenVMS [1]. But nobody ever showed up to do the work, and the last such discussion was in 2011. Given that, I really doubt that there is critical mass to support a port to VMS. It's not enough to just show up with a patch for such a port: there has to be an ongoing commitment to fix problems, run buildfarm animals, and so on, and that takes multiple interested people over a long period. (I think it is pretty much exactly this point that is the stumbling block for the current discussion about whether to reinstate the AIX port [2]: there's nearly zero community enthusiasm about AIX.) Feel free to prove me wrong, but it's going to be a uphill climb. regards, tom lane [1] https://www.postgresql.org/search/?m=1&q=openvms+port&l=&d=-1&s=d [2] https://www.postgresql.org/message-id/flat/CY5PR11MB63928CC05906F27FB10D74D0FD322@CY5PR11MB6392.namprd11.prod.outlook.com
I used to work in Ingres support and one of the things was teaching Architecture of Ingres( seminar so just theory without handson ) .Coming from a trainer( dong tech support training on site consulting etc) people might have seen as shallow my discussions about advanced features of VMS.
I will try to start working on this and probably note the features which justify
the port with some comparative findings) and see if that could get more people interested.
On Sat, Jan 25, 2025, 03:39 Tom Lane <tgl@sss.pgh.pa.us> wrote:
Bruce Momjian <bruce@momjian.us> writes:
> On Fri, Jan 24, 2025 at 01:07:56AM -0800, Christophe Pettus wrote:
>> So, basically, if you want a maintained VMS port, you need to either drive the project yourself, or find others who will.
> This email thread from 2003 says VMS probably doesn't work anymore
> because of lack of testers:
> https://www.postgresql.org/message-id/flat/200301071531.h07FVWI08147%40candle.pha.pa.us#0dbc1439f51ec7842125fb8ae200b6da
I doubt we ever had a working VMS port. There are precisely zero
references to VMS in our commit log, so certainly there was never one
that got removed. It's barely possible that PG "just worked" without
any patches under their POSIX emulation layer, but I could not find
any indication of successful users of PG-on-VMS in the mail list
archives either.
What I did find was occasional suggestions that we port to OpenVMS [1].
But nobody ever showed up to do the work, and the last such discussion
was in 2011.
Given that, I really doubt that there is critical mass to support
a port to VMS. It's not enough to just show up with a patch for
such a port: there has to be an ongoing commitment to fix problems,
run buildfarm animals, and so on, and that takes multiple interested
people over a long period. (I think it is pretty much exactly this
point that is the stumbling block for the current discussion about
whether to reinstate the AIX port [2]: there's nearly zero community
enthusiasm about AIX.)
Feel free to prove me wrong, but it's going to be a uphill climb.
regards, tom lane
[1] https://www.postgresql.org/search/?m=1&q=openvms+port&l=&d=-1&s=d
[2] https://www.postgresql.org/message-id/flat/CY5PR11MB63928CC05906F27FB10D74D0FD322@CY5PR11MB6392.namprd11.prod.outlook.com
On 1/24/25 15:21, Vish Penmetsa wrote: > I used to work in Ingres support and one of the things was teaching > Architecture of Ingres( seminar so just theory without handson ) .Coming > from a trainer( dong tech support training on site consulting etc) > people might have seen as shallow my discussions about advanced features > of VMS. > > I will try to start working on this and probably note the features which > justify > the port with some comparative findings) and see if that could get more > people interested. > I think it would be more useful to establish first whether there are existing users of VMS that would like to see a port of Postgres. What you are proposing, trying to get people to adopt a new to them OS as way to get a port is quite a hurdle. -- Adrian Klaver adrian.klaver@aklaver.com
On 1/25/25 18:36, Vish Penmetsa wrote: > The people who have responded are - Bruce,Christophe,Adrian,Tom,Daniel > and I hope one of you can volunteer to track my mails in the initial stages > 1) Just to be clear I have no knowledge of VMS. Also, to be honest I am not interested in learning it. 2) Per Tom Lanes post there has been no discernible interest in a VMS port of Postgres. Until such time as there is some critical mass pushing for a port I don't your efforts going anywhere. Personally I think if the interest was there it would have manifested itself already. 3) My suggestion would be to carry your project on as blog posts submitted here: https://planet.postgresql.org/ Then you could see if the a fore mentioned critical mass can be reached. -- Adrian Klaver adrian.klaver@aklaver.com
Thanks for the advice. Most appreciated. I was myself very apprehensive about my mails.
On Sun, Jan 26, 2025 at 11:02 AM Adrian Klaver <adrian.klaver@aklaver.com> wrote:
On 1/25/25 18:36, Vish Penmetsa wrote:
> The people who have responded are - Bruce,Christophe,Adrian,Tom,Daniel
> and I hope one of you can volunteer to track my mails in the initial stages
>
1) Just to be clear I have no knowledge of VMS. Also, to be honest I am
not interested in learning it.
2) Per Tom Lanes post there has been no discernible interest in a VMS
port of Postgres. Until such time as there is some critical mass pushing
for a port I don't your efforts going anywhere. Personally I think if
the interest was there it would have manifested itself already.
3) My suggestion would be to carry your project on as blog posts
submitted here:
https://planet.postgresql.org/
Then you could see if the a fore mentioned critical mass can be reached.
--
Adrian Klaver
adrian.klaver@aklaver.com
Unix is a programmer's operating system and obviously programmes like it.
About VMS there is a bit of history that Bill Gates from Microsft had hired Bill Cutler from the VMS Team to build Windows operating system.
And obviously at that time the focus was more on Windows interfaces rather than the VMS like OS which was more of a server.
And when it comes to PostgreSQl whe will be really loking at the server side of things rather than it's user interface capabilities where one migth Windows to be far more
useful and easy.
DEC had lost of the following Industry leading Advantages - DECNet -> TCP/IP
OSF - Linux
Alpha architecture - Everyone caught up with 64 bit chips while DEC Alpha was the first 64 bit chip.
Cluster Architecture -> Oracle & RAC & Unix
And in all these we have'n t looked at the core strengths of how the operating system was built for server operations.
Probably one has to start with the assumption that it is somehow extremely good and try your best to prove it.
One example I could give as to why I am keen on VMS is something about the OS learning about exeutable images to make the startup fast.
Then coming to teh user accounts there is something abour working sets for tuning and these were advatages when you were talking about 32 MB of memory or 64 MB of memory
for large no of users rather than 32 GB or 64 GB for few uers ( I have 32 GB on my Windows PC just for myself)
Then there will be more.
On Mon, Feb 3, 2025 at 3:54 PM Justin Clift <justin@postgresql.org> wrote:
On 2025-01-25 08:09, Tom Lane wrote:
> Bruce Momjian <bruce@momjian.us> writes:
>> On Fri, Jan 24, 2025 at 01:07:56AM -0800, Christophe Pettus wrote:
>>> So, basically, if you want a maintained VMS port, you need to either
>>> drive the project yourself, or find others who will.
>
>> This email thread from 2003 says VMS probably doesn't work anymore
>> because of lack of testers:
>> https://www.postgresql.org/message-id/flat/200301071531.h07FVWI08147%40candle.pha.pa.us#0dbc1439f51ec7842125fb8ae200b6da
>
> I doubt we ever had a working VMS port. There are precisely zero
> references to VMS in our commit log, so certainly there was never one
> that got removed. It's barely possible that PG "just worked" without
> any patches under their POSIX emulation layer, but I could not find
> any indication of successful users of PG-on-VMS in the mail list
> archives either.
As a data point, when the recent incarnation of VMS Software announced
their intention to allow Community sign ups a few years ago, I went
and created an account on their system to investigate. VAX VMX being
one of the first multi-user systems I learned back in the day, before
learning *nix. ;)
Anyway, it went... poorly. Their system is so crap that users can only
have a very specific set of "special" characters allowedin user
passwords:
$#@!%*&
Any other symbols are accepted at password setting time, but actually
cause the user login to fail.
When I attempted to file a bug about this problem, they literally told
me it's not a bug and it working as intended.
So frankly, VMS Software are so completely clueless that I strongly
recommend no-one waste their time and effort on them.
Regards and best wishes,
Justin Clift
Anyway probably it might be good for postgres hackers to go thru some strengths of VMS with an R&D Engineer rather than my speaking - I also worked in presales and I am not an R&D Engineer so...
On Mon, Feb 3, 2025 at 5:17 PM Vish Penmetsa <vish.penmetsa@gmail.com> wrote:
Unix is a programmer's operating system and obviously programmes like it.About VMS there is a bit of history that Bill Gates from Microsft had hired Bill Cutler from the VMS Team to build Windows operating system.And obviously at that time the focus was more on Windows interfaces rather than the VMS like OS which was more of a server.And when it comes to PostgreSQl whe will be really loking at the server side of things rather than it's user interface capabilities where one migth Windows to be far moreuseful and easy.DEC had lost of the following Industry leading Advantages - DECNet -> TCP/IPOSF - LinuxAlpha architecture - Everyone caught up with 64 bit chips while DEC Alpha was the first 64 bit chip.Cluster Architecture -> Oracle & RAC & UnixAnd in all these we have'n t looked at the core strengths of how the operating system was built for server operations.Probably one has to start with the assumption that it is somehow extremely good and try your best to prove it.One example I could give as to why I am keen on VMS is something about the OS learning about exeutable images to make the startup fast.Then coming to teh user accounts there is something abour working sets for tuning and these were advatages when you were talking about 32 MB of memory or 64 MB of memoryfor large no of users rather than 32 GB or 64 GB for few uers ( I have 32 GB on my Windows PC just for myself)Then there will be more.On Mon, Feb 3, 2025 at 3:54 PM Justin Clift <justin@postgresql.org> wrote:On 2025-01-25 08:09, Tom Lane wrote:
> Bruce Momjian <bruce@momjian.us> writes:
>> On Fri, Jan 24, 2025 at 01:07:56AM -0800, Christophe Pettus wrote:
>>> So, basically, if you want a maintained VMS port, you need to either
>>> drive the project yourself, or find others who will.
>
>> This email thread from 2003 says VMS probably doesn't work anymore
>> because of lack of testers:
>> https://www.postgresql.org/message-id/flat/200301071531.h07FVWI08147%40candle.pha.pa.us#0dbc1439f51ec7842125fb8ae200b6da
>
> I doubt we ever had a working VMS port. There are precisely zero
> references to VMS in our commit log, so certainly there was never one
> that got removed. It's barely possible that PG "just worked" without
> any patches under their POSIX emulation layer, but I could not find
> any indication of successful users of PG-on-VMS in the mail list
> archives either.
As a data point, when the recent incarnation of VMS Software announced
their intention to allow Community sign ups a few years ago, I went
and created an account on their system to investigate. VAX VMX being
one of the first multi-user systems I learned back in the day, before
learning *nix. ;)
Anyway, it went... poorly. Their system is so crap that users can only
have a very specific set of "special" characters allowedin user
passwords:
$#@!%*&
Any other symbols are accepted at password setting time, but actually
cause the user login to fail.
When I attempted to file a bug about this problem, they literally told
me it's not a bug and it working as intended.
So frankly, VMS Software are so completely clueless that I strongly
recommend no-one waste their time and effort on them.
Regards and best wishes,
Justin Clift
There was a book called Thriving on Chaos around eighties or nineties which took VMS as an example of Technology Industry Leaderhsip and said their technology leadership might
be taken for granted for only 18 months by which time Competitors could probably catch up.
But then I had said that VMS seems to have become a small userbase operating system not for competitors catching up but sudden boom for small servers where everyone went for Unix
so I felt that it might be worth taking a relook again for postgresql and python and combination.
On Mon, Feb 3, 2025 at 6:34 PM Vish Penmetsa <vish.penmetsa@gmail.com> wrote:
Anyway probably it might be good for postgres hackers to go thru some strengths of VMS with an R&D Engineer rather than my speaking - I also worked in presales and I am not an R&D Engineer so...On Mon, Feb 3, 2025 at 5:17 PM Vish Penmetsa <vish.penmetsa@gmail.com> wrote:Unix is a programmer's operating system and obviously programmes like it.About VMS there is a bit of history that Bill Gates from Microsft had hired Bill Cutler from the VMS Team to build Windows operating system.And obviously at that time the focus was more on Windows interfaces rather than the VMS like OS which was more of a server.And when it comes to PostgreSQl whe will be really loking at the server side of things rather than it's user interface capabilities where one migth Windows to be far moreuseful and easy.DEC had lost of the following Industry leading Advantages - DECNet -> TCP/IPOSF - LinuxAlpha architecture - Everyone caught up with 64 bit chips while DEC Alpha was the first 64 bit chip.Cluster Architecture -> Oracle & RAC & UnixAnd in all these we have'n t looked at the core strengths of how the operating system was built for server operations.Probably one has to start with the assumption that it is somehow extremely good and try your best to prove it.One example I could give as to why I am keen on VMS is something about the OS learning about exeutable images to make the startup fast.Then coming to teh user accounts there is something abour working sets for tuning and these were advatages when you were talking about 32 MB of memory or 64 MB of memoryfor large no of users rather than 32 GB or 64 GB for few uers ( I have 32 GB on my Windows PC just for myself)Then there will be more.On Mon, Feb 3, 2025 at 3:54 PM Justin Clift <justin@postgresql.org> wrote:On 2025-01-25 08:09, Tom Lane wrote:
> Bruce Momjian <bruce@momjian.us> writes:
>> On Fri, Jan 24, 2025 at 01:07:56AM -0800, Christophe Pettus wrote:
>>> So, basically, if you want a maintained VMS port, you need to either
>>> drive the project yourself, or find others who will.
>
>> This email thread from 2003 says VMS probably doesn't work anymore
>> because of lack of testers:
>> https://www.postgresql.org/message-id/flat/200301071531.h07FVWI08147%40candle.pha.pa.us#0dbc1439f51ec7842125fb8ae200b6da
>
> I doubt we ever had a working VMS port. There are precisely zero
> references to VMS in our commit log, so certainly there was never one
> that got removed. It's barely possible that PG "just worked" without
> any patches under their POSIX emulation layer, but I could not find
> any indication of successful users of PG-on-VMS in the mail list
> archives either.
As a data point, when the recent incarnation of VMS Software announced
their intention to allow Community sign ups a few years ago, I went
and created an account on their system to investigate. VAX VMX being
one of the first multi-user systems I learned back in the day, before
learning *nix. ;)
Anyway, it went... poorly. Their system is so crap that users can only
have a very specific set of "special" characters allowedin user
passwords:
$#@!%*&
Any other symbols are accepted at password setting time, but actually
cause the user login to fail.
When I attempted to file a bug about this problem, they literally told
me it's not a bug and it working as intended.
So frankly, VMS Software are so completely clueless that I strongly
recommend no-one waste their time and effort on them.
Regards and best wishes,
Justin Clift
Probably it was wrong of me to send the mail on VMS. I was under some confusion and kind of under some impression that there might have been a port which no one was
using and probably the port might have been outdated until I think Mke or Bruce checked and said that it was never ported on VMS.
On Tue, Feb 4, 2025 at 5:14 PM Vish Penmetsa <vish.penmetsa@gmail.com> wrote:
I have expressed my opinion and if more people agree with you that would be simple. Even without anyone else agreeing with you probably I am not pushing it further.I have posted in VMS forum and if nobody responds it is closed anyway, Anyway I would not call it nostalgia as I was not attached to VMS but Ingres at that time.Anyway I think Python is already available on VMS so that could probably just fulfill my liking for an opportunity to work on VMS again if there is a way.On Tue, Feb 4, 2025 at 3:01 PM Laurenz Albe <laurenz.albe@cybertec.at> wrote:On Tue, 2025-02-04 at 06:55 +0530, Vish Penmetsa wrote:
> I also need to mention that When DEC Alpha servers were released [...]
Perhaps this should go on the pgsql-nostalgy mailing list.
Yours,
Laurenz Albe