Re: [port-srv-reg] "xmp" service type and the unified IANA Service Name and Port Number Registry

[Joe Touch <[e-mail]>]

On 8/25/2011 10:59 PM, Stuart Cheshire wrote:
> On 25 Aug 2011, at 22:13, Joe Hildebrand wrote:
>
>> In case that wasn't just a rhetorical flourish, look up the XMPP federation
>> information for GoogleTalk:
>>
>> $ dig +short -t SRV _xmpp-server._tcp.gmail.com.
>> 5 0 5269 xmpp-server.l.google.com.
>> 20 0 5269 alt3.xmpp-server.l.google.com.
>> 20 0 5269 alt4.xmpp-server.l.google.com.
>> 20 0 5269 alt1.xmpp-server.l.google.com.
>> 20 0 5269 alt2.xmpp-server.l.google.com.
>
> We're delving into software deployment philosophy a bit here, but this is a good example.
>
> A sensible strategy to ship (client or server) software is:
>
>    If an SRV record exists, then use it,
>    else, if a local /etc/services entry exists, then use it,
>    else use the port IANA assigned to you.

I agree.

> This lets you write software and ship it right away, without waiting for some future OS update.
>
> Since the software writer already knows what port IANA assigned, the
> /etc/services entry is only relevant when it's *different* to the port
> IANA assigned. The logical conclusion of this line of thought is that
> it's only necessary to record *exceptions* in your local /etc/services
> file, since all the apps already know their IANA-assigned default (and
> if every OS update were to overwrite your /etc/services file with the
> latest IANA version, overwriting your local changes, you'd get pretty
> annoyed by that).

That's a fair point, except that /etc/services is *also* already used by 
a variety of software as a shared ".h" file, where even the defaults are 
looked up by name rather than coded in by number.

Which means there's no particular harm in having an /etc/services that 
includes the current defaults. It would result in the same response 
using the algorithm above.

It's also not uncommon for sysops to grep /etc/services to find out 
what's running. The common "netstat" command consults /etc/services for 
this purpose as well.

(it'd be useful to know what the source to Apple's netstat uses to 
identify protocols that it didn't know about when it was compiled)

Joe