Re: threading

["Robert Holmgren" holmgren@xxxxxxxx]

** Reply to message from Flash  on Sun, 25 Jun 2006 09:47:52
+0200


> Now, if Carl is right, it should be possible to copy-paste references
> into a header and insert oneself into a thread, regardless what is
> typed into the 'subject' line.

I don't know about "pasting" into the header (I doubt *very* much that you can
do that with any high level Email client, because it almost certainly guards &
protects the header); but if you control the way the mail is generated at a low
level, then you can put anything you want in the header, and it can be as
quirky and complex as you want. I wrote, from scratch, sendmail/getmail
clients that are more manual than Carl's combo (which is to say, Totally Manual
-- barebones implementations of RFCs 821 and 2554) and mine have X-Discontent
lines and private jokes -- which, of course, nobody notices because they aren't
able to see, or don't know how to see, or aren't interested in, the header
("From" and "Subject" aside). The goal of mine was total portability: carry
it around on a diskette, and consult mail from anywhere -- no browser, no
nothing, just a cross-platform command line. Compose outgoing mail with
XyWrite or Notepad, and send. Turned out to be a bit bulkier than I expected,
slightly larger than a diskette due to the need for a sockets DLL, but ... fun
anyway.

An incidental insight of my clients is that the entirety of low-level
handshaking and communication is revealed -- you learn how Email actually works
-- and I was tickled to realize that there is an almost secret network among
the SysOps who design or control mail servers. Out of boredom perhaps, they do
very humorous things. For example, Harvard's mail server speaks in a Romanized
transliteration of Japanese. Others yell crude epithets at remote machines.
All of this goes on in background, and almost no user is aware it exists (you
might see it if your mailer has a "debug" mode). I made early versions of my
getmail/sendmail available here, long ago.

> I don't see any obvious pattern to the message IDs.
> They increment not like TCP sequence numbers (which count bytes of user
> data). Any idea how they increment, and whether it would be possible to
> anticipate the next one in a thread? ('Why would anybody want to DO
> that?' I hear someone ask.

Well, umm, why would you want to do that? All the information you need is
contained, not in future, but in _past_ communications (even if you're stuck
with a really lousy Email client that won't show header info, you can always go
to XySearch, look at a past msg, view the Page Source with your browser, and
read all the header lines). It would be impossible, and pointless, to predict
IDs of future communications. Your own communication will be assigned an ID by
your sendmail server, and if you rely on an ISP's sendmail (don't have your own
mail server), they're going to do the assigning. There's little consistency in
those IDs. Usually (but not always) they contain the date and time, plus some
seemingly random characters.

-----------------------------
Robert Holmgren
holmgren@xxxxxxxx
-----------------------------