Hi Dan,
Thanks for writing up this e-mail, though I'm not sure it contributed much to the discussion. A couple of quick meta-points before I get to the substance: when posting to any public Wikimedia mailing list, please use plaintext, not HTML e-mail and please only quote relevant parts of the e-mail. For example, your reply was not an inline reply to any of Terry's message, so there's absolutely no reason to include it in your reply. Reduce noise and increase signal, please.
Dan Garry wrote:
Proposed solution 1: refine and enable MassMessage (with the view that it will be superseded at some point by a better solution) then work on the better solution.
Yes, do this.
The first thing I observe about these solutions is that both of them recognise that MassMessage is, relatively speaking, a short term solution. We all know it's going to be replaced by something better in the future.
Quick anecdote: https://en.wikipedia.org/wiki/Special:Permalink/301004485#Archiving_bug.3F
This is a wikitext talk page thread from June 2009 about LiquidThreads. I suggested that it might not make sense to invest in a new bot framework for a wikitext talk page archive bot because LiquidThreads was actively being developed and would soon make wikitext archive bots obsolete. :-)
- "How much of a performance problem is EdwardsBot?".
Here's my general thinking. It should help you make a decision.
After over 370,000 edits to Wikimedia wikis, I'm tired of running EdwardsBot. Rather than doing the irresponsible and dickish thing (simply turning off the bot), I filed a bug in March 2012 about finding ways to replace the current system, particularly as global message delivery was being increasingly used for Wikimedia Foundation messages to all Wikimedia wikis (cf. https://bugs.wikimedia.org/35306).
Earlier this year, a volunteer developer, with my help, wrote and implemented a much better system: it now has a proper user interface, it has form input validation, it doesn't rely on Python + cron, it can be deployed to all Wikimedia wikis easily (it scales), it can be easily localized, it ties in with the current architecture really well, etc.
I helped create a better system while maintaining the current system. I did the responsible thing. This is why I have no problem saying to you now that on or about November 1, 2013, I'm no longer maintaining EdwardsBot. This roughly coincides with the date that the MassMessage MediaWiki extension is supposed to be deployed to all Wikimedia wikis. I'm giving the Wikimedia Foundation a set of well-tested (though hackish) Python scripts (https://github.com/mzmcbride/delivery-bots) in addition to a stable, reviewed, implemented, and deployed PHP MediaWiki extension that the Wikimedia Foundation can choose to use or not. I don't care. All of the code is released either into the public domain or under a free license. If you want to continue operating global message delivery and per-wiki mass message delivery with the Python scripts, have at it. I'd even be glad to send over EdwardsBot's password if you'd like to continue using the same user account. A better tool (MassMessage) is now available and unlike Flow, it actually exists and has been reviewed and deployed to Wikimedia wikis. Unlike LiquidThreads, it's not abandonware. But it's entirely your call how you choose to proceed.
A quick comment on the long term solutions. I disagree with the idea that echo should be used actually deliver newsletters (e.g. The Wikipedia Signpost, Wikiproject newsletters). It should notify one of the delivery, not actually deliver. I think, in terms of the future (i.e. Flow), it doesn't make sense to have a bot (whether it's EdwardsBot or a MassMessage bot) starting a discussion on a Flow page to deliver a message to you. That's also doing it backwards.
While I appreciate the time that you and Terry have put in to writing these mailing list posts, I don't really appreciate that you two attempt to speak with authority on a subject about which you have no involvement or experience. I wrote the local bot, I wrote the global bot, I gave a presentation about broadcasting "movement communications" at Wikimania 2012, I created the bug to discuss this issue, and I followed through by helping to build a better replacement. You'll have to forgive me if I don't put a whole lot of weight behind the opinions being offered here. :-)
I'll be the first to admit that the overall concept here is stupid. It's similar to delivering newspapers door to door when we now have the Internet. However, users, thousands and thousands of them, like getting messages delivered to their talk page (or a noticeboard) on their home wiki. Rather than continuing to rely on a hackish Python script, I think we (Wikimedia) can gladly accept the free labor that's being offered to develop this extension and perhaps say thank you to the volunteer developer (Legoktm), rather than writing long, uninformed e-mails about how it doesn't integrate well with software that doesn't yet exist.
My several cents. Over and out.
MZMcBride