As someone who followed the recent syslog-protocol series right from the
beginning and as someone who has implemented RFC 3195,
syslog-transport-tls, syslog-transport-udp and syslog-protocol, and as
someone whose implementations of said drafts are being used in practice
by actual people... I think we have reached a good-enough series of
drafts to publish.

Sure, there are a number of imperfections and some things could be more
precisely specified. HOWEVER, we can always do a second revision. What
we have right now is good enough for most needs and flexible enough to
be extended. There are also a lot of folks waiting on the completion of
this work.

I strongly propose that we do a -transport-tls-13 based on Joe's recent
text proposal. I'd appreciate if that -13 would remove the MUST on
ipAddress. But even if it does not, one can live with it.

Once -13 is published, I suggest we continue to work on 3195bis, which
is already posted, solves a lot of issues but needs a bit of polishing.
Based on our desires, we may also do a lot of other things. 

PLEASE LET US FINISH THE CURRENT WORK *NOW*.

I myself have already departed from the standards way by implementing my
own proprietary protocol, and more and more people will follow the
non-standards route if we do not deliver anything. Given the silence of
almost all other implementers on this list, I think I am not the only
one who did so. Remember that the first draft of the current series was
published in 2003, which was five years ago. If we wait another five
years, I think there will no longer be a need for what we specify,
because everybody will use something else (just think of the widespread
use of proprietary TLS protected plain tcp syslog because we did not
manage to get a document out when there was demand for it...). Netconf
notifications, for example, would probably not have been necessary if
there would have been a decent syslog standard 2 years ago (and we could
have been finished 3 (!) years ago if we hadn't blown the nearly
finished doc set in a meeting for reasons I still find hard to
understand).

Rainer

> -----Original Message-----
> From: Joseph Salowey (jsalowey) [mailto:[EMAIL PROTECTED]
> Sent: Thursday, June 05, 2008 7:49 AM
> To: tom.petch; Rainer Gerhards
> Cc: syslog
> Subject: RE: [Syslog] -transport-tls references to "matching rules"
> 
> Hi Tom,
> 
> I think it would be useful to have recommendations for generic
> application over TLS. I don't think all applications would be the
same,
> but I think there could be some common guidelines.  I don't think we
> should hold up the TLS syslog draft for this.
> 
> Joe
> 
> > -----Original Message-----
> > From: tom.petch [mailto:[EMAIL PROTECTED]
> > Sent: Monday, June 02, 2008 1:15 AM
> > To: Rainer Gerhards; Joseph Salowey (jsalowey)
> > Cc: syslog
> > Subject: Re: [Syslog] -transport-tls references to "matching rules"
> >
> > Replying to myself, and apologies to those who got my first,
> > mangled attempt at this,
> >
> > I have just read
> >
> > draft-ietf-netconf-tls-02.txt
> >
> > which covers an almost identical  territory to transport-tls
> > but with server and client roles reversed; well worth a read.
> >
> > Where we used to refer to RFC2818, it refers to RFC4642
> > (which itself considers relays).
> >
> > It has a reference to RFC5280 and specifies section 6 for
> > certificate paths.
> >
> > It does not consider fingerprinting.
> >
> > It does not consider alternatives to hostname.
> >
> > <rant>
> > As I have said before, I see a crying need for a generic
> > 'application over  ...' I-D for others -  like me - to draw
> > on and reference, lest we keep inventing our (less than
> > round?) wheels.
> > </rant>
> >
> 
> 
> > Tom Petch
> > >
> > > ----- Original Message -----
> >
> >
_______________________________________________
Syslog mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/syslog

Reply via email to