Re: #1047 Path field delimiters and syntax - status

[kaih@xxxxxxxxxxxxxxxxx (Kai Henningsen)]

rra@xxxxxxxxxxxx (Russ Allbery)  wrote on 23.08.05 in <87u0hgaafu.fsf@xxxxxxxxxxxxxxxxxxxxx>:

> Charles Lindsey <chl@xxxxxxxxxxxxxxxx> writes:
> > Russ Allbery <rra@xxxxxxxxxxxx> writes:
>
> >> You have to scan the entire article header anyway in order to verify
> >> the syntactical correctness of the article, and it's simple to note the
> >> location of the Path header as you pass it.  Putting it first really
> >> wouldn't save any noticable time in practice.
>
> > Not if you are a relaying agent. All USEPRO says is that it SHOULD check
> > for the presence of all the mandatory headers.
>
> *NNTP* says that you can't send malformed junk, and verifying that what
> you received is not malformed junk requires scanning the entire headers.
> It's interesting (and rather eye-opening) to hear that UUCP-only sites
> apparently don't have to check the syntax of messages, but it's not
> particularly relevant to the real world.

I don't see how that would work, anyway - as someone currently doing some  
UUCP and previously doing all-UUCP.

It's true you can ship files from A to B through C without C doing any  
kind of analysis of the file, but that means that C can't do anything with  
the file except sending it on.

That is, an intermediate site behaves pretty much exactly like a router  
does with TCP. That router doesn't (usually) do any article syntax check,  
either.

But if you're doing news propagation and the intermediate site is supposed  
to see the article, it goes through rnews - which, in the case of INN,  
does NNTP, I believe, though not necessarily over IP, and thus certainly  
checks. I expect C news did similar stuff.

As for mail, where this middle man stuff need not apply, the behaviour is  
essentially the same as with SMTP is you put the "real" message inside a  
message/rfc822 part - it won't be touched either. That is, the UUCP has a  
different kind of envelope/contents separation than SMTP, and thus is  
interested in different things being ok ... and of course, the original  
UUCP rmail version, just like the original Unix mail version, did not  
include any kind of MTA at all.

These days, any UUCP mail I do is BSMTP and goes directly to the MTA, with  
all the usual checks. (Well, "rsmtp" is what UUCP actually calls it.)

> I really don't see any useful purpose served by optimizing the article for
> faster processing by lazy UUCP-only sites.

I don't see how that'd work anyway.

MfG Kai