On Tuesday, January 7, 2003, at 05:01 PM, Stefano Mazzocchi wrote:
Greg Stein wrote:On Tue, Jan 07, 2003 at 11:53:52AM -0500, Ben Hyde wrote:...Bah. Use SubWiki, check out the Wiki pages into a working copy, make all
It turns out if you build a event driven mail based notification system you shortly there after discover that it's too painful to use. The Wiki model results in editors writing changes so they can preview; and in tangles of nodes going thru periods of total chaos[1] as an editor attempts to make any changes that involve more than one node. Both these make the change stream very hard to read. I suspect that much of that could be resolved by sending mail with changes "when the dust settles".
your changes, then commit them. Regular commit email sends the full bunch of
changes.
Simple as that :-)
Oh, and maybe that's not under the GPL...
You are stating that:
0) download a working copy [this is done only once] 1) go to a page 2) edit it 3) save it 4) commit the page
is comparably simple with
1) go to a page 2) edit it 3) save it
Well duh, it's that what's cool about a wiki. I'm not that interested in revisiting that insight, nor do I think Greg is; instead I'm a greedy bastard and I think we want/need/will-be-much-happier(tm) with both models.
[Ok maybe I'll indulge in a little revisiting that insight... You don't even make the dialectic as stark as it really is. The three step model requires few, if any, tools or skills folks don't already have; while the 4 step model requires CVS, commit rights, etc. etc.]
What I'm trying to discuss is how to get a readable stream of deltas presented to volunteer proof readers. I think that's important.
I've noticed so far that:
- The RSS feed doesn't present the deltas. It appears that events are getting lost.
- That an event driven email feed is too fine grain because it's almost like watching people type. This is a huge problem, I seem to be unable to edit a wiki page correctly in a single try - it usually takes me 3-5. I doubt any proof-reader could stand that!
- It would be nice to have a way to make transactions over large numbers of nodes so
the proof-readers see a transaction rather than the chaos of every flick of the pen.
- That's really helpful when large transformation to the Wiki are done.
- Possible solution: you can get part way to that end by putting some hysteresis into the edit-event -> proof-reading-mail pipeline.
- Thats analogous, and hence not too bizarre, to background build daemons that I've built and I suspect many others have seen or built.
- It would be nice to have a CVS like checkout so editors could make changes locally. There are any number of large edits people might do if they can experiment on them privately.
Greg Stein wrote:
In no way did I say it was "comparably simple" to standard Wiki editing. Of
course not... jeez, just how small do you think my brain is? :-)
Well my brain got a lot smaller after I cut my hair, so bear that in mind.
- ben
