[lug] thanks for the help

[Nate Duehr]

I typoed the CC to BLUG...

-----

Thanks to all on all three lists who had input on the file transfer 
questions I had yesterday!  

It was interesting how many people have all run into similar situations. 

To answer a few of the questions: 

- No, I don't have control of the upload filename.  I liked the comments 
that said uploading to a temporary name and then "renaming" made a pretty 
clean setup.  That idea gave me some ideas about a completely different 
project where I *do* have access to both sides of the file transfer.  

- Bear's comments about sftp'ing /dev/null -- you're just evil, Bear.  
Eeeeevil.  :-)  I like!  Now to find out if sftp plays nicely with 
quotas.  ;-)

- A couple of people were confused.  I did say "sftp" not "ftp"... in 
other words, it's an ssh server with logins turned off (and chrooted 
even...).  The warnings about cleartext user/pass with ftp are of course, 
correct -- if I were using ftp.  (GRIN)

Ultimately I went with the "if it hasn't been accessed in five minutes, 
it's as here as it's ever going to be".  This is based off of some 
assumptions of the uploader's network connectivity and cluefulness (and 
the fact that many are doing it by hand, so they'd see a failed file 
transfer as it happened.)

Luckily also, a little research and I found that the back-end process that 
handles this data can handle a re-upload if the original gets trashed, 
without making a mess... but of course I'd prefer to avoid that!  ;-)

The technique I used was just to mess around with GNU "find" and the -amin 
switch option (amongst others).  It's not 100% fool-proof, but it mimics 
another system we have doing similar stuff, so anyone who'd have to 
maintain it if I weren't here would be able to follow the directory 
trees, etc.

-- 
Nate Duehr, nate at natetech.com