Crash report procedure

[Tereza Snyder]

On Aug 30, 2006, at 4:41 AM, Sivakatirswami wrote:
...

>
> Of  course it means we might need  to build in a little  more error  
> checking,
> perhaps a checksum() on some stack files  and  if  it's not in a  
> reasonable
> range, then  assume the  files are corrupt and delete  them and get  
> fresh copies.
> I need to do better at logging the  context and  scenarios.
> Frankly this kind of thing is a bit over my head...
>

A possible strategy could be to provide, at your server, a small file  
with an MD5 digest or other check associated with each stack. Your  
application obtains that file first, then downloads the stack as  
binary data (perhaps writing it to disk with a "~" prefix, as  
Metacard does for stack saves), then checks the file for correctness,  
and finally renames the file and opens it only if it passes the test.

I know it's obvious, but I had to stick my finger in,


t


-- 
Tereza Snyder

    Califex Software, Inc.
    www.califexsoftware.com