Re: [Gnu-arch-users] Re: situations where cached revisions are not so good

[Tom Lord]

    > From: Miles Bader <address@hidden>

    > I typically update from Tom's sources after reading an email message
    > saying `just added feature X,' so the long latency of a cron-updated
    > mirror is a burden (usually what would happen is, I'd do an update, see
    > nothing changed, be puzzled for a split-second, and then go "oh yeah"
    > and update my local mirror and try again).  

Just as a point of comparison: I have a habit of interacting with
remote archives that always amounts to `push-mirror' and then do
stuff.   I have suitable -SOURCE mirrors registered to make that work
pretty smoothly.  (In the case of _your_ sources, I also sometimes
throw a library-add in there, too.)


    > It also turned out that the actual amount of data transfer was
    > _more_ with the mirror than with a direct connection, because I
    > typically only do a few operations; with a mirror, I either had
    > to remember to use a very specific limit for the mirroring
    > command (and worry about what it should be -- it depends on what
    > changed), but the direct connection basically transfers exactly
    > what I need.

That's my bad.  Mostly, you really should want all the changes in my
archive.  I shouldn't have thrown scm--devo in that archive though.

(You're just as bad in this regard --- I can't afford to mirror your
emacs foo yet :-)


    > I think mirrors (local and otherwise) are a useful part of tla's big-
    > ad-hoc-bag-of-caching-methods, but they're nowhere near the universal
    > solution that some people seem to be implying they are...

Please read my Big Honkin' Message in this thread sent earlier today,
especially the section "Are Good Taste and Complex Heuristic Searching
Compatible?".   It speaks to that issue.

-t