Postgres Pro
Downloads
Home   >   mailing lists

Thread: External Large objects what became of them

External Large objects what became of them

From
"Willis, Ian (Ento, Canberra)"
Date:
Does anyone know what the status of external large objects is?
Did they disappear without a trace and will they ever make a comeback? I was
reading about Xdelta the other night and I thought that postgresql would be
a great interface for this sort of program if it still supported the
interface.

Re: External Large objects what became of them

From
Tom Lane
Date:
"Willis, Ian (Ento, Canberra)" <Ian.Willis@ento.csiro.au> writes:
> Does anyone know what the status of external large objects is?

Not sure what you consider an "external" large object is.  7.0.* has
a few remnants of what apparently used to be support for three or four
kinds of large objects, but the code was nowhere close to working
even if it weren't #define'd out.  I ripped the remnants out for 7.1;
all that's left is the "standard" kind of LO.  The "standard" kind has
been reimplemented a lot more efficiently, courtesy of Denis Perchine,
if that makes you feel any better...

            regards, tom lane

RE: External Large objects what became of them

From
"Willis, Ian (Ento, Canberra)"
Date:
By external large objects I mean storing the objects as ordinary files
through the database interface, giving you a filesystem that supports
tranactions though a postgresql interface. This was one of the features
Stonebraker thought would be very useful. I was reading about xdelta and
remembering the feature being in postgres many moons ago and causing lots of
confusion and thought it may be possible to port Xdelta over to using
postgresql as the backend.
http://www.xcf.berkeley.edu/~jmacd/xdelta.html


-----Original Message-----
From: Tom Lane [mailto:tgl@sss.pgh.pa.us]
Sent: Monday, 20 November 2000 4:42 PM
To: Willis, Ian (Ento, Canberra)
Cc: pgsql-general@postgresql.org
Subject: Re: [GENERAL] External Large objects what became of them


"Willis, Ian (Ento, Canberra)" <Ian.Willis@ento.csiro.au> writes:
> Does anyone know what the status of external large objects is?

Not sure what you consider an "external" large object is.  7.0.* has
a few remnants of what apparently used to be support for three or four
kinds of large objects, but the code was nowhere close to working
even if it weren't #define'd out.  I ripped the remnants out for 7.1;
all that's left is the "standard" kind of LO.  The "standard" kind has
been reimplemented a lot more efficiently, courtesy of Denis Perchine,
if that makes you feel any better...

            regards, tom lane

Re: External Large objects what became of them

From
Tom Lane
Date:
"Willis, Ian (Ento, Canberra)" <Ian.Willis@ento.csiro.au> writes:
> By external large objects I mean storing the objects as ordinary files
> through the database interface, giving you a filesystem that supports
> tranactions though a postgresql interface.

Huh?  LOs stored as ordinary files would by definition *NOT* have
transactional semantics --- aborted changes wouldn't roll back, and
there'd be no way to prevent people from seeing intermediate states
of a transaction, either.

            regards, tom lane

Re: External Large objects what became of them

From
Tom Lane
Date:
"Willis, Ian (Ento, Canberra)" <Ian.Willis@ento.csiro.au> writes:
> http://www.xcf.berkeley.edu/~jmacd/xdelta.html

Ah, thanks for the pointer.  I didn't know what you were talking about
before (a Google search had proven only that "xdelta" is a popular
project name...)

As far as I can tell from this, Xdelta wants to sit on top of a
database, not underneath one.  So I fail to see why it cares what the
large object implementation technology is, or even whether the DB
uses such things at all.

            regards, tom lane