Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 11 Feb 2000 06:40:49 -0600
From:      Richard Wackerbarth <rkw@dataplex.net>
To:        Tony Maher <Tony.Maher@eBioinformatics.com>, <freebsd-ports@freebsd.org>
Subject:   Re: partial cvs
Message-ID:  <00021105173600.00576@localhost.localdomain>
In-Reply-To: <200002110325.OAA04178@shad.internal.en-bio>
References:  <200002110325.OAA04178@shad.internal.en-bio>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 10 Feb 2000, Tony Maher wrote:

> just been following discussion re partial CVS repository on ports.

Please understand that much of my discussion of CVS repository applies to
the main tree more than just ports.

For "ports", I see a way to handle it with the existing tools.

Assume that we divide the universe into "ports developers" and "ports users".

The developers would commit to the present tree structure (perhaps with minor
modification).

Operationally, the users would have no reason to access that tree. There is
little reason to replicate it world-wide.

Instead, a "digest" is generated and distributed. This digest could be simply a
'shar'ed file for each port. I would actually do it a little differently, but
this representation is adequate for this part of the discussion.

Analysis:
Operational characteristics--
CVSup uses the incremental nature of RCS deltas to gain efficiency. This
would not be lost. Adding a line to an underlying file would isomorphically
add a line to the digest. The 'diff'erence between two digests would be just
the 'diff'erence(s) in the underlying file.
Older history in the cvs repository of the digests can be periodically
purged because it can be regenerated from the underlying tree.
The short term history which is useful to CVSup and some users is still present
in the digest repository.

 Cost -- The master ports repository (or its slave) would have an additional
copy of the recent ports tree. Commits would require additional processing to
update the visable "digest" version as well as the underlying tree
representation. However, this only happens on one machine. 

Savings -- The distributed ports tree would be smaller and contain fewer
elements. The worldwide distribution system would be more efficient.



> 
> I had been thinking previously about having a split CVS repository
> "old" and "current" where current is the last 12 months.
> The thing that started me thinking was I wondered  when CVS repository
> would exceed size of a cdrom (I mean an easily mounted cdrom filesystem).
> 
> After reading your discussion, a couple of thoughts:
> 
> it should be a simple to set up a secondary cvs tree which has old revisions
> deleted (-oRANGE).
> Then people who want the cut-down version could cvsup against it
> rather than the full version which could reduce traffic and disk space
> requirments.  This may satisfy large numbers of people.
> And as mentioned the really old info could be obtained from CDROM
> (which I would presume is rare)
> 
> Still leaves problem of how to update two master CVS repositories and you cant 
> outrange via dates (AFAICT).   But could do it on tags introduced at 12
> month (nominal) intervals.
> 
> BTW I no longer keep CVS repository up-to-date and cvsweb supplies my
> 	limited needs ;-)
>     (but am looking forward to getting cvs repository with 4.0 CD release!)
> 
> tonym
-- 
Richard Wackerbarth
rkw@Dataplex.NET



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ports" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?00021105173600.00576>