From: Jose Kahan (jose.kahan@w3.org)
Date: Mon Aug 20 2001 - 11:37:49 CDT
Hello Daniel and other folks,
I've spent time analysing the code in parse.c and file.c, as it has
changed from the last time I looked into it. Some functions are not
commented, related to regexp searches :-/
On Wed, Aug 15, 2001 at 07:58:58AM +0200, Daniel Stenberg wrote:
>
> > My current idea is to use some kind of MD5 hash of the msgid with the
> > received date to name the files. Collision risk should be small (I
> > hope!).
>
> Would be minimal, but won't the file names get annoyingly long and weirdly
> named?
I think that filenames should be small. Some of my work colleagues would
like to use the whole msgid for the filename. I'm not sure this is the
best thing. What do you think?
> The numbers also serve a purpose when reading an existing HTML archive.
> Hypermail can easily just read all HTML files, as it knows their names all
> the way until there are no more files.
I noticed that we read the files in find_max_msgnum(), using a call
to opendir. Regardless of the feature I want to introduce, this seems
a bit dangerous, as hypermail assumes that the only thing found in the
directory are its own messages. Why not? This could always be true. I was
wondering why we don't try open one of the index files that has a list of
all the messages and put there the latest message number? Or just count
them from the file structure.
This would also simplify what I would like to add as a feature :)
I should have a better description of the changes I will like to do this
week, if not the changes themselves :)
-jose