The Plan, part 1

[Roger Dingledine <arma@mit.edu>]

We've got some really good ideas now in terms of what issues we want to
cover for the survey.

It seems like these are the tasks left for us:

1 Decide what user profile questions we want to ask
  * Figure out what the various user types will be initially

2 Finish the OS questions
  * figure out which questions we don't actually need to ask
  * get them ordered correctly, possibly with different orders or
    amounts for different user types
  * word them smoothly, so the end-user can understand them and
    know what we're actually trying to ask (I'll post more about
    this one tonight; it's important.)

3 Do the technical (programming) part of it
  * Make cgi-bin scripts to generate the survey based on how we want
    the questions to look, their order, etc
  * Integrate a mysql db with that so we can sort and collate easily
 
4 Test it:
  * Put it up on the seul page, tell seul-project, seul-dev, seul-pub,
    seul-seg about it, ask them to fill it out and make comments
    wherever they see fit. We should get perhaps one or two dozen
    good people to test it from there. seul-seg consists of non-linux
    scientists for the most part, so that will be a different
    perspective.
  * Take their comments, and fix the survey appropriately. Make sure
    the technical side works really well -- I want to be able to do
    various db lookups to answer statistics questions on the fly, so
    anybody who's curious can do that too on the website.
  * If we're not happy yet, do another round of asking seul-* people
    to fill it out. Ask seul-announce too if the other seul lists are
    being lame. (seul-announce has 430 people who are amazingly
    response.) And then another round of fixing.

5 Actually put it out for consumption.
  * Maybe clear the database, depending on how much things have changed
  * Tell seul-*, freshmeat, slashdot, lwn, cola
  * Get links to it put on relevant webpages (eg LDP)

6 Wing it from there. I'm not about to try to predict it that far.
  But we'll want to consider moving being the linux community, eg
  posting a polite reference to a windows newsgroup (ugh). (Ok, so
  maybe 5 and 6 are the same step...)

Now, 1, 2, and 3 can be done in parallel. The reason why I've been
splitting them up so far is that we're a little bit short on people
resources, and I didn't want to spread things too thinly.

But now that we've got some good progress going...does anybody want
to start working on 3? We can afford to focus on 3 for a little while,
if that's what it takes.

I think we can get away with merging the "OS questions" with the
rest of the questions in the draft that Pete posted. That will mean
that the survey consists of:
1) Pick your why-you-use-a-computer
[cgi script generates appropriate survey]
2) fill in user profile info (age, income, which OS's you like, etc)
3) go down the column of "how important is" questions, checking one of
   four options for each question.
4) optionally write in the "why didn't you ask me foo?!" section and
   similar sections
5) submit and look at statistics.

In message <Pine.SOL.3.96.980708135409.7026A-100000@kanto.cc.jyu.fi>, hvirtane@cc.jyu.fi writes:
>One problem with the survey is that it canot be too big or long (no-one 
>wants to read or go through a long survey) and on the other
>hand it should be able to gather information from very different users.
>That's why I was thinking that somehow it might be possible to start it
>that way, that some kind of users might be saved of reading through some
>parts of it and some of other parts of it. For eample it is important to
>gather quite specific information from wordprocessor-users, if they really
>insist WYSIWYG processors or not and if they want their processors being
>able to read all the formats of the processors and so on... but people who
>never use wordprocessors at all could be saved of reading through these
>questions totally.     

So the survey questions (apart from the user profile section) are a set
of common questions plus the extra questions for that "user type", though
these extra questions might be interspersed with the common questions
however we like.

The fact that we're presenting the questions in a different order depending
on the user type means that it will be less meaningful to compare answers
between different user types, but I'm willing to live with that.

We want to make certain that the whole survey is on one page. The person
taking the survey needs to know how long the survey is, as well as be able
to go back at any point and change answers. I don't want to deal with
cookies.

>So a clever typology of tasks is important. Technically for example
>html-language hypertext will do the job, if the keywords are cleverly
>planned so that a reader of the survey can understand what is behind each
>keyword. 
>
>hv

Please argue with me. I may sound like I'm making declarations of What Will
Be, but i've been up for way too long, and I may well be spouting nonsense
instead. :)

--Roger