Win95: "Corrupt intallation detected." - cause known, solution needed

[Tim Peters]

[Bengt Richter]
> It wouldn't be that hard to gin up a little stand-alone .exe to do
> what md5sum.py does.

[Tim]
> Thinking-cap time:  the problem doesn't come up at all unless
> they're having trouble downloading .exe files.

[Bengt]
> 01-04-17  20:17  6,272,831 D:\Python\Python-2_1.exe   (my last download)
> 00-07-17  15:45     32,768 C:\UTIL\lsm.exe            (my ls md5 utility)
>
> if p() is probabability of download error for n bytes (even on
> the same day ;-),
> what is p(6,272,831):p(32,768) ?
>
> Not good enough odds? Or were you thinking of the subset of
> problems unaffected by byte length (yet perhaps checkable with a
>length number) ?

p is usually so close to 0 or 1 that you can't tell the difference, and is
the latter when they're managing to use a download method that
systematically corrupts LF bytes into CRLF pairs on Windows -- or vice
versa.  Both change length.

Keep the context in mind!  This is Windows, so that particular form of
download problem is depressingly common, and these are generally newbies.
"MD5" doesn't mean anything to them.  I've "dealt with" hundreds of these
reports.

If you want to make a standalone .exe available, that's great, and I'll
gladly refer non-newbies to it the next time there's a downloading problem
nobody can figure out (so please post a URL).  The length check almost
always settles it, though; switching to a different download method then
almost always solves it; and I never see such reports from non-Windows
downloaders.

the-conditional-probability-of-most-interest-is-the-one-intersecting-
    "newbie"-and-"windows"-ly y'rs  - tim