I have just committed to CVS the changes to the way Ethereal handles captures.
Now captures are saved to a file unknown to the user, that is, to a
scratch area. Ethereal calls tempnam(NULL, "ether") to do this. So, the file
starts with the name "ether" and is put into the system-defined temporary
directory. (On Solaris 2.6, `grep P_tmpdir /usr/include/stdio.h` shows
/var/tmp). The directory can be overridden by the user by defining the TMPDIR
environment variable before starting ethereal.
The user can use File | Save to save the trace to a file of his chosing.
The user can also use File | Save as to save a trace file with another name.
When Ethereal exits, it makes sure to delete the temporary capture file, if
one exists.
I think this is a more intuitive way of capturing packets. The user can
continue to capture and re-capture until he finds a trace good enough for
saving.
I added the list of authors to doc/ethereal.pod (yes, another file to
maintain). I put the info about setting TMPDIR into the *.pod file and
recreated the man page.
And I updated my e-mail address. I changed jobs. :) (and yes, I made sure IBM
knows about ethereal and wiretap as previous works, so there won't be any
intellectual property problems)
--gilbert
