Re: 1.7.1 release date?

[Christopher Faylor <cgf-use-the-mailinglist-please at cygwin dot com>]

On Sun, Dec 06, 2009 at 01:07:28AM +0000, Dave Korn wrote:
>Christopher Faylor wrote:
>>People *will* be seeing problems when they click on the "Install Cygwin
>>Now".  If I was more clueless, I'd see them myself, since I just type
>>"http://cygwin.com/setup.exe"; in internet explorer when I want to
>>upgrade.  So, I would be blithely downloading a new version of cygwin
>>which could possibly end up confusing me.
>
>>And, for the record, I don't care about any problems a directory rename
>>might cause any private mirrors, including, coincidentally enough, my
>>own private mirror.
>
>I don't think those two arguments go together.  They're both about
>extrapolating from one's personal experience, but it seems one says
>it's ok to do so, and the other says it is not.

The point that I was trying to make was an extension of this:

http://cygwin.com/ml/cygwin-developers/2009-12/msg00062.html

Although I didn't really make that clear in any way.

I was not trying to say that my usage was typical even though it isn't
likely that I am the only person who does things that way.  If I am one
of M people in the whole world who types "http://cygwin.com/setup.exe";
to run setup are you going to say it's ok if they are suprised by the
upgrade?  If so, why is it ok for those M people and not for the N
people who run old copies of setup.exe directly from their systems and
ignore warning messages?  Do you think N is significantly higher than M?

Putting that aside, the point I was *really* trying to make was that if
it is a given that people will be surprised and damaged by the upgrade
then it is very likely that people who just click on "Install Cygwin
Now!" will be affected.  You can't make a case that there aren't a lot
of people who just go directly to the Cygwin web site and click on a
link without paying attention.  We *know* that people do that.

So, I would submit that, if there are problems in the release, we'll see
them a lot more from people who download from the web than we will from
people have their own copy of setup.exe.  If one accepts the premise
that updating will cause problems and one accepts the premise that
people will do blind clicking, then I think that it would make sense to
fix the problem generally.  The general problem would not be solved by
directory naming.  It would only, AFAICT, be handled by something like
your suggestion:

http://cygwin.com/ml/cygwin-developers/2009-12/msg00056.html

To reiterate, if I were to concede the point that upgrading would cause
hearbreak then I would want to have a general solution to the problem
rather than a "fix" which causes anyone who runs an old version of
setup.exe to, at best, get a somewhat inexplicable error message or, at
worst (if the release directory was not renamed), just silently get no
update at all.

>These are both valid ways to draw up arguments with the backing of
>experience behind them, but I think we often don't remember that we (by
>which I mean all us involved in this particular thread) are the
>*minority* of cygwin users in the world,

Aren't we here because Corinna suggested that there are likely other
users like her who just, against the recommendations of the website, run
setup.exe directly from disk?  I was, actually, trying to point out what
I think is a more common use case.

>and our usage patterns and comfort zones are not necessarily those
>which apply to the majority of cygwin users.  (As evidence, I point to
>various bugs in setup.exe that have existed for some time, but none of
>us ever knew about because we already had existing cygwin
>installations, and never tried doing a fresh one from scratch during
>that period.  Without having done a survey, if I had to guess, I'd
>guess that we operate actually *very* unlike how most others use their
>cygwin installations.)

Are you sure you want to use that argument?  That would seem to suggest
that Corinna's originally noted problem isn't that big a deal since
"normal" people don't keep old cygwin installations around.

It seems like what you're saying is that setup.exe lacks good testing.
It should be a well-known fact that the project lacks testers.  I
routinely ask for volunteers up to keep the Cygwin test suite alive.  QA
isn't a fun job so no one wants to do it.  Testing obviously isn't a
panacea but, since we have very little, a lot of regrettably fixable
stuff slips by.

Also, I think you presume too much to say "none of us ever knew".
There's a difference between "not knowing" and "not doing" although the
basis for both problems is usually pretty much the same in a free
software project.

cgf