There was no JSON or other nonsense. No pages upon pages of "API documentation". A user could figure it out in a matter of minutes.
This was an easy way to retrieve chunks of a mailing list, from circa 2002 onwards, in a format that was easy to work with, using only basic UNIX utilities.
It was like dumping a mail spool.
Given the Sirens of gratuitous complexity that never stop singing on today's web, I fear for the future of this project.
But the point of this comment is to say thanks to the author of Gmane v1. Great project, great execution.
Yeah, I still don't know why Lars wouldn't let them have the existing front-end code. I understand that maybe it wasn't pretty under the hood, but who cares? It worked great for many years and was extremely useful. It could always be cleaned up and refactored a bit at a time.
Now they're having to reinvent the wheel. And while I shan't presume to tell them how to do it (especially since it's at their expense), I wonder if the infrastructure they're cooking up will be as future-proof. It certainly won't be as simple to reimplement if Gmane changes hands again someday, and it won't be conducive to setting up other instances of Gmane on a smaller scale.
Any chance of donating a full dump of all the data to https://archive.org to preserve this treasure trove for future generations? Imagine if the dejanews newsgroup archives had not ended up in Google's horrific JS interface but open for anyone.
A 'side-effect' of Firefox's reader mode is it renders news sites that rely on Javascript to display a 'please subscribe before reading this article' overlay readable...};-)
I wonder why they'd choose to mix Python and PHP and not just pick one language for EVERYTHING. Seems like you'd risk having to implement some functionality twice.
Suprisingly, not that large. My archive of NNTP messages has about 10M messages in only 5.6GB of space, with no compression. SMTP messages accumulate far more header cruft (in particular, chatty Received headers and often times DKIM as well in practice), so I'd expect 10M SMTP messages to be about 2-5 times as large as the same number of NNTP messages.
The total size of gmane is only on the order of a few TB at best. I think the author said that it was only 2TB--for 14 years of history!
I almost feel the same way: I love reading mailing lists and RSS feeds with Gnus via NNTP.
But the Web interface to mailing lists and groups is also extremely useful. Nothing like clicking a Gmane link in LWN and being able to instantly browse the thread like a newsgroup.
http://gmane.org/lists.php <-- a simple gzipped list of all the groups and their corresponding gmane category prefixes
ftp -4o lists.gz http://gmane.org/lists.php
zgrep and a little bit of sed was all that was needed.
http://download.gmane.org/gmane.[listname]/[start]/[end] <-- bulk download list messages in the range [start] to [end]; no brackets needed of course.
There was no JSON or other nonsense. No pages upon pages of "API documentation". A user could figure it out in a matter of minutes.
This was an easy way to retrieve chunks of a mailing list, from circa 2002 onwards, in a format that was easy to work with, using only basic UNIX utilities.
It was like dumping a mail spool.
Given the Sirens of gratuitous complexity that never stop singing on today's web, I fear for the future of this project.
But the point of this comment is to say thanks to the author of Gmane v1. Great project, great execution.