At 10:36 AM +0200 9/19/03, Christian Limpach wrote:
>"William Allen Simpson" <wsimpson@greendragon.com> wrote:
>
>>  Then, you have 1,000 (10,000? 100,000?) folks like me waiting about 14
>>  hours for `cvs up`, compiling -current, discovering that the build fails,
>>  re-cvs (only about 7 hours this time), -u build, discover the flist is
>>  bad, re-cvs, -u distribution (still running at this moment)....
>>  Basically, 2+ days, and still not updated, and this on a dedicated
>>  test machine.  I shudder to think what production systems are doing.

So you ignored the instructions to cd into the src/crypto/dist/ssh 
directory before doing your 'cvs up' and 'make USETOOLS=no && make 
USETOOLS=no install' then?

>If you do not have a checked out source tree to start from, you can grab
>tarballs of the source tree from our ftp servers.

Or you can grab the individual files (not in tar form) directly from 
the ftp server. They did turn off the ability to tar-up directories 
on the fly (too expensive on the current servers), so you'd need to 
create the directories & do mget *.

>  > Perry (and others) say there are some ideas about the future, and I'm
>>  interested in helping make that happen.  The status quo is not good.
>
>I'd hope that at some point we'd be able to provide binary updates which
>include only the updated files and can be extracted over an installed
>system.

Too many architectures, too many updates. Between 1.6.0 and now (on 
release), there have been hundreds of little changes pulled up at 
various times. While you could only do the ones with security 
advisories about them, you'd be missing the other fixes to the base 
system. This really needn't ever be done for -current IMO (we're 
discussing this on current-users)... -current users take 
responsibility for their own systems and should be able to update 
parts (or the whole) as necessary.

Just my thoughts,
Mike
-- 
Bikers don't *DO* taglines.