Re: svn commit: r169092 - /spamassassin/trunk/spamd/PROTOCOL

[Justin Mason <jm...@jmason.org>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


Daniel Quinlan writes:
> Or, we could switch the order around a bit:
> 
>    the classic report as: ASSERT CLASS=SPAM LOCAL=SET REMOTE=SET
>    the classic revoke as: ASSERT CLASS=HAM LOCAL=SET REMOTE=REMOVE

I like that.  +1

BTW, what about just "TRAIN" or "CLASSIFY" instead of "ASSERT"?  I get
what you mean by ASSERT, but that verb in general has implications of
code-quality, not user interaction -- it's a very generic term.

In particular I'd prefer to use a word that indicates that a user is
asserting something about the *class* of a *message*, not just that
they're asserting something in general.

On the other issue -- the ASK interface --

regarding ditching REPORT, CHECK, PROCESS, etc. in a future rev -- I
really don't like that idea.  there's actually very little hard work
involved in taking the verb from the first protocol line, and doing the
little bit of conditional code for each of those, and *removing* verbs
from a TCP protocol is a *very big deal* -- like moving from POP2 to POP3,
the name of the protocol changes, kind of big.  I don't think any verbs
have ever been removed from HTTP.

however if we ever want to do future variants of CHECK, I'd be fine with
using "switches" on the "command line" as we're proposing doing above,
using an ASK verb, instead of adding new verbs.

- --j.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Exmh CVS

iD8DBQFCfV35MJF5cimLx9ARAvLfAKCHSVMfrc9tp9EANwfLtKaUz7LEvwCgm2j+
PHwamfhKxIOt1wyd1qFmIV0=
=cPeJ
-----END PGP SIGNATURE-----

Re: svn commit: r169092 - /spamassassin/trunk/spamd/PROTOCOL

[Daniel Quinlan <qu...@pathname.com>]

jm@jmason.org (Justin Mason) writes:

> BTW, what about just "TRAIN" or "CLASSIFY" instead of "ASSERT"?  I get
> what you mean by ASSERT, but that verb in general has implications of
> code-quality, not user interaction -- it's a very generic term.

How about DECLARE?

TRAIN and CLASSIFY imply a learning database and a corpus, respectively.
I'm also thinking about "Class" (see Michael's post) vs. some other
name.  We might want to be able to set other properties or flags in
addition to classes.

> In particular I'd prefer to use a word that indicates that a user is
> asserting something about the *class* of a *message*, not just that
> they're asserting something in general.

Hmmm... I see what you mean, however, it's potentially not just the
class.

To jump off on a bit of a tangent, I wonder how we would support a
"unsubscribe" function (like a lot of people are asking of AOL and other
ISPs that have a "report as spam" button which some dolts use as an
unsubscribe).  Maybe ASSERT (or whatever) class=unwanted, who=sender.

> [...] like moving from POP2 to POP3, the name of the protocol changes,
> kind of big.  I don't think any verbs have ever been removed from
> HTTP. [...]

I'm fine with being conservative, hence my desire to get this right now,
but it's a stretch to compare us to POP2 to POP3.  A better comparison
would be some other intra-product protocol.

Daniel

-- 
Daniel Quinlan
http://www.pathname.com/~quinlan/