Horton
Menu

Thread

  1. Re: Re: 7.2 items

    Hannu Krosing <hannu@tm.ee> — 2001-05-14T05:30:45Z

    Franck Martin wrote:
> 
> I think OID should be truly unique in the world as to make it easier for
> replication. If OID are real unique number (not in a table, not in a
> database, but in the world) then replication can be easily built with
> OIDs...
> 

Exactly! That is what the Mariposa project did - they made OIDs uniqe
and 
consisting of 32bit site id + 32bit local OID. I guess this could be
split 
some other way too, like 20 bit site id + 44bit local or any other.

IMHO the best would be a scheme of 32bit site id + 32bit local, but each 
site can get additional site ids from some central (for a supersite)
table 
when it sees that it is near runnig out of oids.

-----------------------
Hannu

  2. RE: Re: 7.2 items

    Franck Martin <franck@sopac.org> — 2001-05-14T05:34:02Z

    I think OID should be truly unique in the world as to make it easier for
replication. If OID are real unique number (not in a table, not in a
database, but in the world) then replication can be easily built with
OIDs...

Cheers.

Franck Martin
Network and Database Development Officer
SOPAC South Pacific Applied Geoscience Commission
Fiji
E-mail: franck@sopac.org <mailto:franck@sopac.org> 
Web site: http://www.sopac.org/
<http://www.sopac.org/> Support FMaps: http://fmaps.sourceforge.net/
<http://fmaps.sourceforge.net/> 

This e-mail is intended for its addresses only. Do not forward this e-mail
without approval. The views expressed in this e-mail may not be necessarily
the views of SOPAC.



-----Original Message-----
From: Lincoln Yeoh [mailto:lyeoh@pop.jaring.my]
Sent: Monday, 14 May 2001 3:45 
To: Bruce Momjian; PostgreSQL-development
Subject: [HACKERS] Re: 7.2 items


At 01:20 PM 10-05-2001 -0400, Bruce Momjian wrote:
>Here is a small list of big TODO items.  I was wondering which ones
>people were thinking about for 7.2?
>
>---------------------------------------------------------------------------



4) Not really important to me but can serial be a proper type or something
so that drop table will drop the linked sequence as well? 
Maybe:
 serial = old serial for compatibility
 serial4 = new serial
 serial8 = new serial using bigint
(OK so 2 billion is big, but...)

5) How will the various rollovers be handled e.g. OID, TID etc? What
happens if OIDs are not unique? As things get faster and bigger a nonunique
OID in a table might just happen.

Cheerio,
Link.


---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@postgresql.org so that your
message can get through to the mailing list cleanly

  3. Re: Re: 7.2 items

    Franck Martin <franck@sopac.org> — 2001-05-14T11:01:23Z

    Hannu Krosing wrote:

> Franck Martin wrote:
> >
> > I think OID should be truly unique in the world as to make it easier for
> > replication. If OID are real unique number (not in a table, not in a
> > database, but in the world) then replication can be easily built with
> > OIDs...
> >
>
> Exactly! That is what the Mariposa project did - they made OIDs uniqe
> and
> consisting of 32bit site id + 32bit local OID. I guess this could be
> split
> some other way too, like 20 bit site id + 44bit local or any other.
>
> IMHO the best would be a scheme of 32bit site id + 32bit local, but each
> site can get additional site ids from some central (for a supersite)
> table
> when it sees that it is near runnig out of oids.
>
> -----------------------
> Hannu

As I'm thinking about it there is a utility called uuidgen which generates
such numbers.

On my Mandrake distro it is part of the e2fsprogs package. Orbit uses it to
generate unique numbers too.

--------------
The  uuidgen  program creates a new universally unique identifier (UUID)
using the libuuid(3)
       library.  The new UUID can reasonably be considered unique among all
UUIDs  created  on  the
       local system, and among UUIDs created on other systems in the past and
in the future.

       There  are two types of UUID's which uuidgen can generate: time-based
UUID's and random-based
       UUID's.  By default uuidgen will generate a random-based UUID if a
high-quality random number
       generator  is  present.  Otherwise, it will chose a time-based UUID.
It is possible to force
       the generation of one of these two UUID types by using the -r or -t
options.

 The   UUID   of   the   form   1b4e28ba-2fa1-11d2-883f-b9a761bde3fb   (in
printf(3)  format
       "%08x-%04x-%04x-%04x-%012x") is output to the standard output.
-------------------

Cheers.
Franck@sopac.org

  4. Re: Re: 7.2 items

    Tom Lane <tgl@sss.pgh.pa.us> — 2001-05-14T14:39:34Z

    Franck Martin <franck@sopac.org> writes:
> The  uuidgen  program creates a new universally unique identifier (UUID)
> using the libuuid(3)
>        library.  The new UUID can reasonably be considered unique among all
> UUIDs  created  on  the
>        local system, and among UUIDs created on other systems in the past and
> in the future.

"Reasonably considered"?

In other words, this is a 64-bit random number generator.  Sorry, I
think the odds of collision would be uncomfortably high if we were to
use such a thing for OIDs ... certainly so on installations that are
worried about running out of 32-bit OIDs.  It sounds to me like uuidgen
is built on the assumption that only relatively small numbers of IDs
will be demanded from it.

			regards, tom lane

  5. Re: Re: 7.2 items

    Gilles Darold <gilles@darold.net> — 2001-05-14T15:03:18Z

    Franck Martin wrote:

> I think OID should be truly unique in the world as to make it easier for
> replication. If OID are real unique number (not in a table, not in a
> database, but in the world) then replication can be easily built with
> OIDs...

The Apache server has a UNIQUE_ID implementation and it is really
unique in the world, I use it for my web apps. Their implementation
is really simple an works fine. It is 19 alphanumeric bytes long.

Regards,

  6. Re: Re: 7.2 items

    Ian Lance Taylor <ian@airs.com> — 2001-05-14T17:21:03Z

    Tom Lane <tgl@sss.pgh.pa.us> writes:

> Franck Martin <franck@sopac.org> writes:
> > The  uuidgen  program creates a new universally unique identifier (UUID)
> > using the libuuid(3)
> >        library.  The new UUID can reasonably be considered unique among all
> > UUIDs  created  on  the
> >        local system, and among UUIDs created on other systems in the past and
> > in the future.
> 
> "Reasonably considered"?
> 
> In other words, this is a 64-bit random number generator.  Sorry, I
> think the odds of collision would be uncomfortably high if we were to
> use such a thing for OIDs ... certainly so on installations that are
> worried about running out of 32-bit OIDs.  It sounds to me like uuidgen
> is built on the assumption that only relatively small numbers of IDs
> will be demanded from it.

uuidgen with the -t option generates a UUID which includes the current
time and the Ethernet hardware address.  The value is about as
globally unique as it is possible to create in 128 bits.  The same
algorithm is used by DCE, and a variant is used by DCOM.  To be used
properly, you need to coordinate on one machine to ensure that
different processes on that machine don't generate the same UUID.

Here is a description:
    http://www.opengroup.org/onlinepubs/9629399/apdxa.htm#tagcjh_20

Ian