Subject: Re: digest
To: Hubert Feyrer <>
From: Alistair Crooks <>
List: tech-pkg
Date: 03/21/2001 14:29:08
On Wed, Mar 21, 2001 at 02:33:30PM +0100, Hubert Feyrer wrote:
> On Wed, 21 Mar 2001, Alistair Crooks wrote:
> > I think we all get the picture by now that you don't like digest.
> The point is not digest, but the whole implementation. I've mentioned
> several of the flaws before, but didn't even get an answer on it so far.

You have written a lot of things up until now.

> > So, let's assume for a moment that we ditch digest, and move to a
> > modified cksum(1) utility.
> Using one external utility over another one is the smallest problem here.
> To recap, some of the problems that I see with the current implementation
> are:
>  * the handling of the digest pkg bypasses the normal dependency handling
>    at large, where it could be used. Simply depending on digest>=xxxxx
>    would be enough, yet the wheel was reinvented there.

No wheel was re-invented - the dependency was done in the same way as
the pkg_install tools, another basic building block.  As Dan has
mentioned, the libtool dependency is a good reason why we shouldn't do
things with digest>=xxxxx.  I think you're looking at pkgsrc as being
a single mass.  I urge you to view it as a number of layers, with
digest right at the bottom, pkg_install slightly above it, then
lukemftp and bmake, and, finally, the rest.

>  * Having all the code in an external utility (outside pkg_*) prevents
>    us from using the new digest formats inside the +CONTENT files. I'm
>    quite surprised that this was not considered to be worth fixing when
>    this went in in the first step. 

You deleted the part of my original message that talked about
bootstrapping the digest utility.  If the digest checks were to be
rolled into pkg_install, how are you going to check the patches and
the distfile checksum for pkg_install?

And to answer your concerns, when I'm through with answering rational
mail on this subject, I shall add that functionality to pkg_install.

>  * And finally, it was claimed that cksum(1) was too hard to modify to
>    add new algorithms. I think that's proved wrong now. My point is, if
>    you had actually bothered discussing things and give some time for
>    testing, we could have avoided a whole lot of mess.

I claimed it was too difficult to modify, yes.  From my dealings with
it, it was.  And, to repeat myself from the original mail, it's
dynamically linked, it's in the wrong directory, it needs a version
flag, and different versions exist on foreign operating systems at the

You, and other members of the packages team, were given adequate
opportunity to comment and test.
> So, let's not blame it on the digest pkg, which is the smallest problem
> here.