Hi! As I'm quite fond of github's issue tracker, I'll answer the cons :-)
* markdown parsing makes it hard for novice users to simply copy'n'paste examples, code snippets or anything else that conflicts with markdown
How many "novice" users are there reporting bugs actually? In my experience, no novice users are really reporting bugs, because EVERY issue tracker is way beyond them. But I could be wrong there.
* no way to attach a file to bug reports (this is the most serious thing IMHO)
Yup. You're right there and if this is happening often, it's a show stopper.
* I'm not sure if non-committers are able to attach tags and thus help categorizing tickets
That's true. But: We require registration for adding new issues - so we're not having any "non-commiters" and I'd like to ask, how good - in your experience - are the user-submitted categories actually?
But thinking about it, it is true, that we will loose the two categories Flyspray currently has: an issue type (bug, feature request, todo) and an issue category (ACL, Template, etc.). As long as we're not using tags like "ACL - bug", "ACL - Feature request", etc. (which we will certainly NOT be doing) we will lose that. Again, how important is that?
We don't have a specialized team in which we could auto-assign ACL issues to the ACL team and template issues to the template team (actually, we have a template team. Bad example ;) ).
* we'd need to write an export/import script to move all previous tickets to github (any volunteers?)
If I can find the time, I'll do that.
* we would have ID clashes for the 100 or so existing ticket IDs (from pull requests)
Isn't that what was meant by talking to github about starting at issue ID #5000 or so?
* as far as I know, existing developers are familar and happy with flyspray's features (correct me if I'm wrong)
Familar yes, happy - so so. What I really like about the idea is the combination of the issue tracker, pull requests and the code base.
* there's no easy way to send a weekly bug summary mail to the mailinglist
No easy way, but a way after all (using github's API).
* we rely with a crucial part of our devel process on a 3rd party that proved to be deaf to our support pleas before
But wouldn't that be even worse for the actual code, that is already in "their hands"?
I don't want to convince anybody here. After all, I'm a very small Dokuwiki developer and cannot really say something about the current issue situation. I like github's integrated way of developing and contributing, but after all I can understand Andi's thoughts.
Kind regards Dennis -- DokuWiki mailing list - more info at http://www.dokuwiki.org/mailinglist