Talk:Shaila
From espians
I'm still floored by tav's suggestion.
So... here's some issues that come to my mind:
- Much work for me to provide for those who can't be bothered to read irc logs?
- would they pay me to produce such a "product"
- Much work for me to do by hand every day
- talk of helpful tools have yet to materialize
- Much work with little comepensation
- yeah, I'm not the only one, but for a daily responsibility?
- I don't have London-esque financial responsibilities
- Espians without broadband
- meatspace responsibilities
- sit on your ass for 6 hours then try doing physical labor
- try physical labor all day then 6 hrs of shalation
--jeffarch 02:25, 19 June 2007 (UTC)
I've been thinking about this a bit recently, and I've been wondering: what's the best result out for the least effort in? Even doing the collecting and tagging of threads of conversation is taking you a really long time for a given input conversation, and yet tav's asking you to provide richer summaries. Scary!
So, how about going the other way? Rather than organising (sorting the thematic strands of discussion) and augmenting (tagging the resulting themes), how about distilling (writing brief summaries of each of the themes)? Organising and augmenting solves the problem where people want to look up all that has been said about a particular topic. Distilling solves a slightly different problem, where people want to see a summary of all that's gone on without wading through the logs.
The thing that makes it easier is that whilst you'd still have to tag all of the content, you wouldn't have to actually include any of the content at all. You wouldn't even have to link to it, if you didn't have time.
So, for example, I've been reading the backscroll since last evening and this morning on #esp, and I found that the two main topics of discussion were Superpuddle and Jeffarch Shailaing. To find that out took a bit of work, i.e. reading the logs, but that's something we should all be doing anyway. The shailar can then take the first step just by jotting down "On the night of 18th June 2007, Qopi and GReaper discussed Superpuddle and Jeffarch discussed and wrote about his Shaila Problem that tav created." This already helps people in the present by letting them know whether they should read the backscroll or not given their particular interests, and it helps us in the future to know when certain things were discussed, decisions made, philosophies engineered, and so on.
Next, you can expand on the outline. The idea is that you start with something that's trivially useful, like the very short thematic summary above, and then you build upon it. So you might take out key quotes from the discussion, like "the point to me is to be able to empower more social change organisations by being able to provide easy low cost free software communications solutions; all I'm saying really is that it makes sense for people to pool resources" from Qopi, say. You might provide a single link into the logs about where the discussion took place.
As long as you start with something useful (a brief summary) and then add layers of usefulness (a few quotes, a longer summary, a considered response, a link to the original discussion, a link to other similar discussions, some mention of antecedents, undertones to the discussion, and on and on), you can stop at any time and you still have a useful, finished product. This means that, unlike the organise and augment process where if you stop half way through you end up with a half finished product, you're kinda incrementally saving and you can step out at any time, and someone else who isn't the Grand Exhalted Chief Shailar could add their own bits as long as they understand the process.
I'm not sure what to call the distillation process, if not distillation... accretion? Stratigraphy? Stratigraphical Shailaing! At any rate, I wonder if the idea makes any sense?
--Sbp 09:44, 19 June 2007 (UTC)
I like the distilling idea, I sort of did something last week when I did a bunch of one liners describing various conversations going on. That's not too much hard work and just requires a bit of thought and time. Anything more complex or elaborate, I probably would not take on personally.
--Evangineer 10:15, 19 June 2007 (UTC)
Additional: One thing I've learned from lots of historical research is that you often don't know what bits of information are going to be useful in the future. Ergo, archive everything! But of course you can make informed guesses, especially when archiving everything isn't feasible. The first question you have to ask is, is it feasible to archive everything?
With shailas, what you're trying to do is to archive all the structure that exists around the raw data. We're following the Archive Everything design pattern with respect to the logs, for example, but we're not expressing any metathoughts about the logs. We're not intrepreting swathes of the logs, and talking about the context of the logs. We're not picking out what we think are the most pertinent points. To get the shaila with the most benefit, we just need to do what comes naturally. Hence the layering principle that I wrote about above. (Cf. a whit on this).
--Sbp 16:00, 19 June 2007 (UTC)
+1
Yes, I think the distilling idea is good too - brief meeting minutes and summaries might not be "shaila" but they are easier a lot to do and actually more useful until better shailating tools emerge (if they do). -- Josef/qopi
sbp+jeffarch Shaila Discussion
Here's the log of my chat with Mr. Palmer: --jeffarch 01:45, 22 June 2007 (UTC)
sbp oh! what did you think about my notes on Talk:Shaila by the way?
sbp I don't think we chatted about it yet
- jeffarch looks
jeffarch to me, one should be able to follow threads like "shaila" and "superpuddle" thru the logs
jeffarch thru separate log files
jeffarch like your example today...like wanting to ignore salvage's comments on marketing
jeffarch (but this is the Shaila issue, not TTK)
sbp yeah, I know
sbp but if you just start with the tags then add other stuff...
sbp ...I was just wondering if that might be a quicker way of achieving it
sbp less monolithic, I mean
jeffarch what tags would one start with? ;p
sbp depends what the topic of conversation is
sbp shaila, superpuddle...
jeffarch am thinking that the tag -> link(s) structure is easiest
jeffarch less copy/pasting, rethreading
sbp well... how do you write a shaila now?
sbp you go through the logs and, right
sbp copy and pasting and rethreading, which takes a lot of time
jeffarch yup
sbp I'm thinking more about a three step process
sbp first you make a list of all the themes discussed. no links to where the discussion was or anything like that
jeffarch I'd like to see something like the chat log with an extra column
sbp ooh, that'd be neat
sbp annotating the logs
sbp collaborative anno... NOOooOo@
sbp all of this is so much work, that's the problem
sbp I'm thinking about what the *most* important bit is, and starting with that
sbp hence the "On the night of 18th June 2007, Qopi and GReaper discussed Superpuddle and Jeffarch discussed and wrote about his Shaila Problem that tav created." thing
sbp it's the way that you build the shaila up that I'm thinking about
sbp not so much the nature of the shaila itself
jeffarch yeah, only one person should have access to the special log screen
sbp "start small and add layers", if you want a slogan out of it
sbp special log screen?
jeffarch with the extra column
jeffarch that'd have a selection of tags, or type-in one that's not there
jeffarch with some method to determine tag start/stop ranges
jeffarch that way each line wouldn't have to be annotated
- sbp does $ touch patterns/SlogansAreEnough.txt patterns/AlwaysUsefulLayers.txt
sbp hmm!
sbp AlwaysUsefulLayers is how I should freaking model design patterns too!
sbp what a dunderhead
sbp what a revelation!
sbp damn, these are two awesome design patterns
sbp TAKE HEED, VERILY, MR. ARCHAMBEAULT! :-)
jeffarch heheheh
sbp so my problem here is that I like to write about design patterns, which are basically... damn where did I put that summary...
jeffarch yeah, where is that?
sbp [[[
sbp The architect Christopher Alexander had created them in A Pattern Language (1977) as a method of formally documenting solutions to design problems. The idea is that if you're having to solve the same thing over and over again but in slightly different ways, if you give a name to the solution and document and understand it thoroughly, you'll save much repetition of effort.
sbp ]]]
sbp it's there
sbp now, one of the problems I have is that when I write up design patterns I tend to be so lazy that I only put down the name
sbp which is terrible really, but actually it's much better than not even writing down the name
jeffarch as long as a definition for the name is elsewhere
sbp hitherto, I've had the idea that it's best to fully document design patterns when you get round to them
jeffarch then, where it occurs would be nice, too
sbp but thanks to my rant on Talk:Shaila I've now come up with that AlwaysUsefulLayers pattern--i.e. the idea that you should do the most useful thing of minimal effort first and then add layers so that you always have a finished product
sbp and when you apply that to my documentation process of design patterns itself, you get the conclusion that I should reject only documenting design patterns in full. just a quick note here and there is good, if it's layered with utility
sbp yeah, sometimes the name isn't really enough
sbp I actually forget what some of the design patterns that I scribble down by name only really meant. heh
sbp but... close enough
sbp bit of leakage around the sides is fine
jeffarch if there are enuf examples of the name, you can use those to rebuild the def
sbp the main idea is that you're making it topic prominent
sbp ooh, good point
sbp a design pattern that rots away possibly wasn't worth it anyway
sbp poor thing didn't make it. its time wasn't right!
jeffarch my "worry" is that a 2-liner of "what happened in IRC" just isn't very useful
sbp ah, but it's more useful than nothing
sbp it's also more useful having five summaries than one in-detail garnering from the logs, I propose
sbp again, it's not that you *can't* add more useful stuff
sbp just that you should start with the little things and then build up
sbp it's not a thought about the content of the shailas, it's a thought about the way you make them, the process by which you construct them
jeffarch but by tagging the lines, such summaries can be generated on-the-fly
sbp but who's going to do the tagging?
jeffarch the shailar ;)
sbp okay. and quis custodiet ipsos custodes?
sbp (who's going to police the police?)
sbp it's too much work, I think
jeffarch alot less work than it is now
sbp you're welcome to do it that way, I don't mind! I just thought the problem was that you've got too much work
sbp really? less work tagging the logs directly? hmm
jeffarch less than pasting links into another document, even
sbp fair enough! I'm happy... I got two damn good design patterns out of this :-)
sbp got an idea for constructing a log-tagging system?
jeffarch bonus! :)
jeffarch well, I'm thinking auto-completion, but that means I'd have to know most of the tags, so a list would be nice, too
jeffarch (thinking how to rig it into the current log system, with zope)
jeffarch storing the tags is't a big issue, I can use zpysql, or whatever that built-in sql thing is
jeffarch I'd prolly wanna work off a trimmed version of the logs, too
jeffarch ie, no joins/parts, no bot traffic, etc
jeffarch then the current log viewer could have a select list for <all> or any of the tags used that day
jeffarch "who" is already distinct, so who+tag=summary
- sbp writes this up here: http://inamidst.com/patterns/DesignPatterns
- sbp reads what you just said
sbp what about a dropdown list?
sbp it'd be nice to be able to tag whole ranges of lines too
jeffarch ooo "hooks to a fractal"
jeffarch I like that
jeffarch ya, I mentioned the whole ranges earlier
sbp whoops
jeffarch jeffarch with some method to determine tag start/stop ranges
jeffarch I like CamelCase too, it's easy to build heirarchies with 'em
sbp aye
jeffarch dropdown list could get messy, but it'd really be there for clues, ie "hmmm...what tag could I use for THAT?"
sbp yeah, tag discovery. makes sense
jeffarch am pondering how to deal with "oh, BTW..." within a tagged range
sbp you could just do some kind of annotation thing in a text file and have a script pull it out
sbp superpuddle{
sbp 03:15 some line...
sbp 03:17 some line...
sbp }
jeffarch I want to get away from text files
sbp wouldn't help with tag discovery though
sbp heh, fair enough
sbp javascript magics? :-)
jeffarch yeah, prolly
jeffarch good ting I bought a js book a few months ago
jeffarch (and have yet to open it)
sbp chuckle
sbp looked at jQuery?
jeffarch nope
jeffarch hmmm...so I'd have to store the tags as well as link|tag
jeffarch t has messed with the log foo...he may be interested in this
sbp it would be cool for filtering
sbp what about multiple tags and tag hierarchies and stuff?
jeffarch KISS ;)
jeffarch but if someone wanted to build it with that, then sweet!
jeffarch tag hierarchies...sounds like part of TTK
jeffarch but I've always said that TTK and shailas can share the same tech
sbp Aperture has it... it's the only tool I know of that does
sbp I don't yet know how it works in practice
jeffarch programming + espra vs marketing + espra
jeffarch cuz I would tag alot of today's log with just 'espra'
jeffarch or with CamelCase, "EspraNetwork" may
sbp yeah, bit single-subject the chat today
