Re: [RFC] Git User's Survey 2010 (resend)

On Sun, 4 July 2010, David Bainbridge wrote:

It sounds like I was revisiting old territory in some cases :-)

Nevertheless the discussion is interesting.

The thoughts about free text analysis were interesting though ...

Sidenote: analysis of free-form questions and of 'other, please specify'
responses was done with the help of program anyway.  But it only did
counting, and you had to categorize by hand.

On 4 July 2010 13:14, Jakub Narebski [off-list ref] wrote:

quoted

On Sun, 4 July 2010, David Bainbridge wrote:

quoted

Have you considered the question: 'How long have you been using Git?'.
This combined the expertise level might yield some useful information,
and could provide some weighting for the 'Getting started' questions
too. There are some problems with the question in that it could be,
say, 2 years intermittent use, or 2 years continuous use, and these
are different, of course.

Well, such question was replaced by the question about one's own
proficiency in Git in 2009 survey (IIRC).

quoted

quoted

=== 06. Rate your own proficiency with Git: ===
(single choice)

 * 1. novice
 * 2. casual, needs advice
 * 3. everyday use
 * 4. can offer advice
 * 5. know it very well

The problem with question "How long have you been using Git?" (beside
what you have mentioned) is that it is hard to answer correctly / in
detail for longer times.

Perhaps something like this:

=== xx. How long have you been using Git? ===
(single choice, with other)

 * less than week
 * between a week and a month
 * a few months
 * a few years
 * duration (please give units)

    * I don't remember

quoted

What do you think?

I think that the granularity is about right. I think we could imagine
where we would expect people to be proficiency-wise with those
periods. Not sure about the last option ... You have everything from
days, to weeks to months to years in the main options. What did you
have in mind?

Scratch last option.  What I had in mind that when people more or less
remember when the started using git, they could write e.g. "7 weeks",
or "5 months".  Let's leave only pre-defined answers... but leave
question about proficiency in git too; this would help distinguish
between continuous use and sporadic (intermittent) use.

quoted

quoted

The demographic information is useful if you want to understand
whether Git is (for example) considered less usable in countries where
English is not the native language. But then if you want to know the
native language of the user rather than where they live then perhaps
this should be asked? (I was born in the UK, but now live in Sweden
... so I am perhaps more conscious of this!). Where you are born does
not necessarily identify your native language either.

The problem with this question in the past was twofold: first, people
sometimes mistok it for question about _programming languages.  Second
should this question be about primary language / languages, or about
language ones prefer the computer to communicate with?

Obviously you have been though this loop before! But perhaps if you
are using drop-down lists then this makes it obvious what is required.
I guess both the programming language, and the preferred interaction
language could be relevant.

We may want to know whether Git is more 'acceptable' to the Java
community or the C/C++ community for example, or whether certain
suggestions for improvement are coming from a community around a
particular language.

I can bring back the question about programming languages, with or
without 'other, please specify' free-form answer.  I wonder about
wording: should it be (as before) about being "proficient" in the
programming language, or about simply "using with git"?  Should it
perhaps be matrix form, asking about proficiency in given programming
language?
 

Then with the human languages this could drive the localization of the
commands, as well as the documentation localization. Interesting ...

Yes, it would be interesting, as the localization is back (!) thanks
to Ævar's work on adding i18n infrastructure to git (C, shell, Perl)
for localizing command messages.

If this question is back, then perhaps also question about what
one would like to have localized: GUI tools (gitk, git-gui - they
are localized, but perhaps not in one's preferred language), command
messages (porcelain), documentation i.e. manpages (here for us is
a question about technology to use), commit message templates
(e.g. "Merge foo into bar", "Revert nnn").  Have I missed something?

By the way, the important question is also which questions to remove,
not only what questions to add.

quoted

quoted

Do you want to separate work usage and 'non-work' usage at all? The
two experiences may be completely different. If I am working at home I
can use what ever I want, and work how I want. This will probably not
be the case in a corporate environment, so I would report one set of
results for 'work' and another for 'non-work'. This could generate two
survey returns for a single user. Is this intended?

Hmmmm... food for though.  It can be mentioned somewhere close to
beginning that one can fill this survey more than once (from different
computers if using default version with cookies) for different roles
one plays.  Perhaps with checkbox "additional role" (aka. "I have already
filed this survey" ;-))...

Thats sounds reasonable ... I think it is probably useful to capture a
user-role response. The priorities of users and their needs may be
completely different in a work situation.

Well, the version with specifying roles might be overly complicated.
I think it would be enough to have checkbox saying

  I have already filled this survey in different role

to catch number of responders as opposed to number of responses.

-- 
Jakub Narebski
Poland