Re: Machine Identity

[Keith Moore <[e-mail]>]

Jeroen Massar wrote:
> Stephane Bortzmeyer wrote:
> [..]
>> There are solutions for some protocols (SSH keys of RFC 4251 or Host
>> Identifiers of HIP in RFC 4423 are two good examples) but no general
>> "identity layer" in the Internet architecture.
> 
> IMHO HIP (RFC4423) is that exact identity layer and additionally solves 
> a number of other issues that it was made for.
> 
> The big problem though is that actually getting programs to use this is 
> cumbersome, simply because a lot of programmers don't know about it.

I like HIP.  But if there is any OS that ships with HIP support, I 
haven't heard about it.  And as soon as we get a "machine identity" I 
suspect we'll realize that what people want to identify isn't exactly a 
"machine" anymore.

> Same goes for the use of SRV records, IMHO, instead of point to AAAA/A 
> records for a service, one should *always* use SRV records.

NO.  SRV records can only be used with applications that are specified 
to use SRV records - otherwise you'll break compatibility with legacy 
applications.  And there are very few of these.

> SRV records solve loadbalancing, failover and a lot of other issues 
> mostly already in DNS.

Yeah, and they create a tussle between DNS operators and users.

> Webbrowsers for instance don't do SRV yet. One of their arguments is 
> 'lookup time', the same reason that some 'webdevelopers' used IP 
> addresses in their websites as that would skip the resolving step...
> 
> This is more about educating programmers than anything else...

Seems like IETFers need education at least as much as programmers do...

Keith