Re: SEUL: Init'l survey stuff - org'n (looooong. Sorry.)

[Roger Dingledine <arma@mit.edu>]

In message <358924DE.8F2B8393@iname.com>, pete_st_onge@iname.com writes:
>[Watching Linux watchers watch Linux...]
>> There's certainly a lot of space for data in there. I wonder if we could
>> make enough use of it for it to be worthwhile.
>     I guess it depends on the question you want to ask. On one hand,
>having some basic info on user activities would be useful. On the other
>hand (tekkie mode: ON), keeping a large ('blind') repository of user
>activities would probably be a boon to further development in that we
>could see via the survey just how intensively linux is being used. If
>some sort of error tracking mechanisn were integrated, it would
>facilitate getting rid of kernel or application bugs, etc. [ie. log
>startup, poll every 120 seconds to open sessions, log shutdown - logs
>without shutdown are crashes, etc.] (tekkie mode: Off)
>     Please set me straight if I'm off the mark, but does this sort of
>resource currently exist? If not, man, it would be useful to build...

I don't know of any resource that does this. [That's one of my major
gripes with the linux world these days -- there are so many things that
nobody knows have been implemented, and so many things that people wonder
have been implemented. I've been trying to fix this, and it would be
really cool if somebody did (I'll provide hardware if they provide
the rest...) ]
So that means we should
a) find out who to ask. A post to redhat-devel and/or debian-devel would
   probably stir up some discussion.
b) start designing it. We should have thought about it more before we make
   such a post. But I don't have enough time to do that right now. :(
   I also think we should focus on one thing at a time.

>     Seems to me though that we're starting to see two avenues opening
>up for the group: the online user survey and the user monitoring
>project. IMO, both of these projects are worth pursuing as they will
>both provide needed information which we couldn't otherwise get. Given
>that both projects seem to fall within the mandate of our group, we
>should be able to pursue both of these.

Correct. We might want to (once we've got a basic design in mind)
post to a couple places and try to find some more people to branch off
and work on the monitor-tool, so we can proceed in parallel. With the
number of people we have right now, though, I think working serially is
the best option.

>[Survey stuff...]
>>>User Info
>>>     Experience with OS, which OS used, how long, what uses
>> OS's. Plural. I'm curious about experience and uses of all sorts of
>> os's -- linux, dos, os/2, win31, win95, winnt, mac, other unix (should
>> we just lump them? probably not quite that much.) User info will also
>> include name, email address (contact info), age, profession, income (?),
>     Yeah, we _need_ to get a good idea of the demographics of potential
>users - it would give an excellent marketing tool to the advocacy
>groups, and would probably be useful to linux-friendly commercial
>software companies.

Apart from providing general demographics information, this also
provides us with data to correlate with the other answers they put:
cad programmers think stability is important, or people who play games
don't care about task switching, or whatever. People say "yeah, but
we already know that. That's intuitive." to a lot of things, but
* I want to document it quantitatively
* I think we'll notice other things that weren't so intuitive

But I think I'm starting to repeat myself.

>     Agreed. I guess we're at the point where one of us should be
>collecting these emails to keep a draft survey list up-to-date and
>available. I don't mind doing it - but I'm starting to think that such
>an animal might better be hosted on a web page rather than just sending
>it via email all the time. 

Correct again. I'll set up the seul-research section of the seul page
further. I've given you (pete) an account, and I'll explain further
details in private mail.

>     I share your wariness about getting too deep into details of
>'uncommon' (ie. something not generally used by the average user)
>programs; I originally saw the db's, languages and system utils as being
>separate optional surveys and wanted to include them for the sake of
>completeness. 

We can easily have them separate surveys: I think if we write it carefully
we can have a very good base survey starting from the rambling bullet list
I gave yesterday. Then we can have a detailed wordprocessor survey, for
instance, or spreadsheet survey. This requires actually going into the
wordprocessor projects and figuring out what they're working on, what they
can do better, what they need help with, what they think is important, etc.
But this is a Really Good Thing to do anyway, it just takes a lot of work.
Somewhere in that effort would also be a documentation project to detail
all the wordprocessor projects going on, their status, etc. From there, we
could coordinate them and try to actually get a couple of them to succeed
in a big way. But I'm creating another project here, aren't I. :)

(see http://www.seul.org/dev/apps/ and
http://independence.dunadan.com/developers/software/ but we're a long way
from having something like I described above...)
 
>> * dependency checking (is there a windows counterpart to this?)
>     Don't rightly know. Then again, I can't say I know exactly what
>dependancy checking is; I'm willing to wager that it's the checking for
>the modules refered to / required by the program while being compiled or
>run?

While it's being compiled (if .tar.gz) or installed (if rpm/deb). (Though
I'm sure there are plenty of other ways to do it.) The basic idea is to make
sure that if a program gets installed, it will do what it says it does,
rather than failing to start because you have the wrong libc installed
elsewhere on the system.

>> (I'm trying to think of windows plusses here too. Somebody who actually
>> likes the thing want to help me out? :)
>     Hey! I resemble that remark!! :-)
>
>     One thing (amongst others :-) NT has that you haven't mentioned for
>linux is a good 'inspection' or diagnostics program (winmsd,
>f'rinstnace) to give you access to all the system information you need
>(hardware / software / IRQ's / DMA's / etc.). Some of this would be
>simple to implement under linux, but (forgive my ignorance - I still am
>learning linux) the seeming lack of centralized repository of system
>information (Registry). Mind you, there are all of the *.rc's, and I
>imagine that even these could be searched for and through for requisite
>information.

Also, there's the /proc hierarchy. You'll find all sorts of great stuff
living in there. It's true though that I haven't seen something similar
to norton sysinfo for linux. I bet there's a KDE version in the works,
though. Probably several completed apps, given the volume of apps they've
been churning out lately. (If not, somebody should mention it to the kde
folks.)

>     Given that we're aiming to make linux attractive to those who are
>not all that familiar with the system, I think it is also important that
>users can find the information they need about their system quickly.
>This may be a useful 'first recommendation' for software for linux in
>general...
>
>
>> (Actually, I should be trying to think of plusses for every OS -- os/2,
>> mac, etc as well.)
> (mac) using a system which protects the user from making potential
>dangerous changes to system configuration

Actually, linux does this, and to an extent nt does this (but it doesn't
really). But it's a good point. It's also something that is more
end-user-worded than "Do you care about PNP?" :)

>-- 
>Pete St. Onge - McGill U.  Limnology - Fun with Ropes & Buckets
>pete_st_onge@iname.com         http://wwp.mirabilis.com/4322052

--Roger