Re: opinion needed on feature/download

Matúš Kukan píše v Út 27. 11. 2012 v 13:34 +0100:
> On 26 November 2012 11:46, Petr Mladek <pmladek@suse.cz> wrote:
> > It was me who come up with the idea or removing md5sum from the file
> > name. Well, it was after Matus came up with the idea of introducing .md5
> > files :-)
>
> Wasn't it David who came up with .md5 files ?
> https://bugs.freedesktop.org/show_bug.cgi?id=56131#c1
> But I think he meant to have them in git, not on web.

I see. The .md5 and even .url file was David's idea and I was not sure
about the real meaning.

> > IMHO, you need not have md5 in git to detect the download corruption. If
> > a file is corrupted, the md5 sum from the .md5 file does not match. Or
> > did I miss something?
>
> But then you would also need to check md5sum of the .md5 file, that's
> recursive problem.
> They need to be in git, otherwise there is no point in having them.
> md5sum could match if they were created that way and passed to you somehow.
> (Unless, I am missing something.)

The question is what is the purpose of the md5sum here:

We either want to check that the file was not corrupted during the
download (network failure). In this case, the .md5 file is enough and we
do not need to have it git. Note that there is no recursion. If the .md5
file is corrupted, md5sum does not match and you just need to download
both files again.

Or we want to make sure that people use the only single version of the
tarballs (security?, preciseness?). In this case, we might need md5sum
in git. But this is pretty non-standart solution. I think that it is too
paranoid and I am not sure if it is worth the effort having the complex
names. For example, if you want to work with the file and do not
remember md5sum, you need to search the directory to be able to write
the right name...

> > As I said, I do not have any strong opinion. If we remove md5sum from
> > the filename, it will cause extra work for distro package maintainers.
> > On the other hand, it might be slightly easier in the long term.
>
> Me neither.
> So, probably does not make sense to discus this unless someone decides
> we want to change it.

Yup.

Best Regards,
Petr