Subject: Re: sup vs cvsup
To: Hisashi T Fujinaka <htodd@twofifty.com>
From: David Maxwell <david@fundy.ca>
List: port-i386
Date: 12/22/1999 14:10:35
On Wed, Dec 22, 1999 at 08:49:57AM -0800, Hisashi T Fujinaka wrote:
> >From what I've read, it sounds like cvsup isn't the ideal solution because
> available for all platforms. Is this right?

There are a couple reasons it is not ideal yet. TNF has said it would not
be an official method for updating NetBSD sources because it is not available
for all (or even close to all) NetBSD platforms. That's actually because the
Modula3 compiler is not available. IMHO, that's a good point of view, both
for users to have native binaries available, and for people running servers
to have native server binaries (for machines which probably run NetBSD).

As a result, the only CVSup server I'm aware of for NetBSD sources, is the
FreeBSD.org one, cvsup.de.freebsd.org. Two disadvantages there - since 
the FreeBSD guys aren't an official NetBSD mirror, I imagine the timeliness
of the sources is more 'stratum 2 - slower to update' than official NetBSD
mirror sites. Also, the FreeBSD server is more heavily loaded, and CVSupd
is a more demanding server process.

Unfortunately, in my latest time trials sup from netbsd.org is faster than
CVSup of NetBSD sources from cvsup.de.freebsd.org. I'm checking into whether
I've made a mistake and I'm not getting the same set of files, whether it's
server load, or something inefficient in the cvsup binary. I'll post real
results for comparison when I have them.

> But with the supscan failing more often, is cvsup the way to go?

Depends on how the above comments affect the way you are using the sources.

> If the supscan fails, will the cvsup fail as well?

I believe cvsup.de.freebsd.org gets its updates with rsync. If their
rsync fails, you won't get any updates. 

The CVSup web pages describe the advantages of the protocol better than
I can. One complaint I hear on these lists from time to time is when tagging
is done, sup replaces every file you're tracking. CVSup will do that in a
much better way.

As for binaries for the other archs, we're proceeding as time and access
to platform allows.

(For those who've gotten the wrong impression of me lately, please remember
I took on the m3 porting work to stop an m3/CVSup/NetBSD flamewar thread.)

-- 
David Maxwell, david@vex.net|david@maxwell.net -->
All this stuff in twice the space would only look half as bad!
					      - me