rpm: alpha vs numeric

Axel Thimm Axel.Thimm at physik.fu-berlin.de
Thu Nov 13 07:51:13 UTC 2003


On Wed, Nov 12, 2003 at 04:28:20PM -0800, Jesse Keating wrote:
> On Wednesday 12 November 2003 01:41, Axel Thimm wrote:
> > o Version your rpms, so that this bug is not triggered. Therefore
> >   a sane versioning scheme not jumping back and forth from alpha to
> >   numeric segments is unevitable. See also the lengthy thread about
> >   it with the disttags for the RH family finally recommended as
> >
> >                     rh7.3 < rh8.0 < rh9 < rhfc1
> 
> Not to start this on yet another list, but I didn't seem to see a
> consensus on the use of "rhfc1" to indicate fedora core 1.  it's
> rather ugly, and causes every RH person I've talked to about it to
> shudder.

OTOH it has been adopted by most non-RH repos, and there was no
comment against it from any RH people official or not IIRC.

> The other method, which seemed much cleaner, was to use "0.7.3" for
> rh73 "0.9" for rh9, and "1" for fc1.

That method was proposed about 3 time VERY LOUD on the -devel list
(with using fdr or whatever distid), but no RH people commented, so it
was dropped or better died after a silent death in an agony of 4 or 6
weeks.

> If you _still_ want text, perhaps "0.rc73" and "fc1", so that we
> don't run into "r" being older than "f".  Using text IMHO is a nono
> and should be avoided at all costs.

"Text" is important. People want to quickly identify the origin and
context of the rpm. If the numeric scheme had been chosen to be
ordered (e.g. numversion_of(RH9) <_rpm numversion_of(FC1), like in the
originally anticipated 7.3 < 8.0 < 9 < 10, or the retrospective
reversioning of RHL to 0.7.3 < 0.8.0 < 0.9 < 1) one could use the same
distid (e.g. "fdr" in one of my late proposals or even "", if you
don't want any nonnumeric characters there), but see above about the
acceptance of these schemes.

So the rh < rhfc distid tag is the best solution avoiding heterogenous
segment comparison (to retun on topic), and providing a natural
boundary for human readers and rpm order parsing the like between the
proper release build number and the distversion. It also fits nicely
into the already established rhX scheme of dist labeling, without
forcing rebuilds of pre-FC1 rpms, and obfuscation of RHL versions with
a zero in front (which seems to be something RH does not want at all
costs).

This has all been presented on -devel, please check the archives.
-- 
Axel.Thimm at physik.fu-berlin.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/fedora-legacy-list/attachments/20031113/4ad320be/attachment.sig>


More information about the fedora-legacy-list mailing list