Re: md5sums

[Ben Gertzfield <che@debian.org>]

>>>>> "Jason" == Jason Gunthorpe <jgg@ualberta.ca> writes:

    Ben> I believe this has been discussed before, and the general
    Ben> consensus was that dpkg is slow enough already as it is;
    Ben> generating md5sums on the fly, while it would be a great
    Ben> thing to have as an option, probably isn't good for everyone
    Ben> all the time, as some people run Debian on extremely slow
    Ben> processors, down to even a 68030 or a 386.

    Jason> Optional perhaps? md5 hashing chews up a minor amount of
    Jason> time compared to the actuall disk io, especially when you
    Jason> do it as the data is flowing into the file you are writing.

Sure, but including a tiny md5sums file in with the package is pretty
simple and requires very little effort on everyone's part.

However, there is a not inconsiderable amount of diskspace used
by md5sums (even as they are) -- I have a bit over a megabyte
(1053K) of md5sums in my /var/lib/dpkg/info directory. Here
are the biggies:

170     tetex-extra.md5sums
113     ncurses-term.md5sums
112     tetex-base.md5sums
51      wnn.md5sums
36      afterstep.md5sums
33      gimp.md5sums

Curse you, tetex!

    Jason> Putting it in dpkg means that automagically evey package
    Jason> suddenly has valid md5sum files.

Optionally, yes. Does dpkg even *have* a configuration file? :) How
would this be controlled?

Personally, what I'm more interested in is having the debsums tools'
functionality built-in to dpkg.

Ben

-- 
Brought to you by the letters R and L and the number 8.
"More testicles means more iron." -- Lunchlady Doris, The Simpsons
Debian GNU/Linux -- where do you want to go tomorrow? http://www.debian.org/
I'm on FurryMUCK as Che, and EFNet and YiffNet IRC as Che_Fox.