[Mailman-Developers] X-Message-ID-Hash header (was Re: Python 3)

[Jeff Breidenbach]

I should probably report the experience of mail-archive.com,
which has supported a homebrew message-id hash algorithm
for over 6 years.

Only one organization has ever used it (LibreOffice).
They put it in the Archived-At headers, not message footers.
However, links did end up getting manually referenced
in message bodies 575 times. Folks here might want to take
a look at how people used it. Christian Lohmaier from
LibreOffice is a good contact if anyone wants to talk to
him directly.

The homebrew has algorithm turned out to be a pretty bad
design. Base-64 got us in trouble. One memorable bug involved
hashes prepended with a minus sign. This caused havoc
with the search engine resolving the links, who interpreted the
minus sign as a negation operator.

Blending the List-Post address into the hash was supposed
to resolve crossposts. In practice it complicated things
enough that LibreOffice incorrectly computed some of the hashes.
There were also times when mail-archive.com wasn't doing
that great a job either of running everything smoothly.

Finally, we really get in trouble when asked to merge multiple
archives, due to a mailing list migrating from server to server.
It happens, and those migrations often change both List-Post
and List-Id. Blending either into the hash is a headache.

So the homebrew hash was dumb idea, and will gracefully
retire in favor of the standard mailman hash algorithm. It
was pretty neat to see some usage, though. And Archived-At
still looks like a winner.

Cheers,
Jeff


Alogithm:
https://www.mail-archive.com/faq.html#listserver

Usage:
https://www.mail-archive.com/search?l=all&q=go.mail-archive.com