[plt-scheme] Version Check

On Apr 27, Anton van Straaten wrote:
> Eli Barzilay wrote:
> > 
> > It showed that it was hardly used, which means more outdated
> > installations.
> 
> But it also implies not much demand for being reminded about
> updates, and it doesn't tell you how many people had deliberately
> "outdated" installations and were quite happy with that.  It doesn't
> say anything about any need that users have.  But no need to argue
> with me about this - I understand the various rationales behind
> encouraging updates.

This is a standard catch 22 thing -- the people that need to check for
updates are exactly the people who will never check for it themselves.

[BTW, for a long time I didn't appreciate the need for this -- I
always forget that with a highschool user base there are always
surprises.  One day I was wondering around the Israeli educational
system since I heard rumors about a functional-programming unit for
schools.  After some digging around I got to a site of the Israeli
"national teachers center for computer science", which is some
ministry-of-education managed thing -- and to my horror, not only do
they talk about version 53 (!), they have their own v53 download, just
to make sure it is always possible to get dusty software...  I just
checked it again now, and it's still there --

  http://cse.proj.ac.il/migvan/fp/DrScheme.htm

The first link on the page is for a v53 installer.  It's working, if
you want to get some nostalgia.  At the bottom of the page they
mention that the newest version is 201, and that it has a problem:
there is no `atom?' function, and go on describing in a separate page
how to add it.]


> > The thing is that "install time" goes away -- the installers are
> > all simple unpack-a-directory things now (except for the extra
> > windows things like the start menu entries).
> 
> In that case, a dialog box should appear when DrScheme is first run,
> before it ever checks for updates.  Subsequent upgrades could
> trigger this dialog box again, if you like, although the answers to
> the questions from the previous run should ideally be retained.

A dialog has the problems I mentioned earlier.  But I think I have a
solution that is acceptable in all ways, another message soon, after a
bit more hacking.


> (E.g. a tiny config file could be devoted to nothing but recording 
> whether the question has been asked, which gets overwritten and thus 
> reset by an upgrade).

(No need for config files -- there's a standard preferences facility.)


> > But install time is a vague concept anyway: consider lab
> > installations (which are exactly the kind of crowd that this is
> > supposed to target).
> 
> I'm not sure what technical issue you're thinking of.

A lab administrator installing PLT and getting a question that is
really intended for the users.


> I have experience with exactly one lab, at a high school, and I
> would think that per-workstation reminders about updates are the
> last thing that the people responsible for a lab would want.  But
> that's not an area I'm very familiar with.

Right -- the lab admin doesn't want it.  When 30 kids get the message,
they will just use their `escape' reflex and nothing bad happens, but
chances are that the lab admin will be told about it.

-- 
          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
                  http://www.barzilay.org/                 Maze is Life!