ietf
[Top] [All Lists]

Re: The ecosystem is moving

2016-05-13 15:00:24
Yes, we could certainly do better.   I think you are mistaken about the
worldwide web, though.  Not only is it an example of a distributed database
of massive scope, but the various additional hops to which you referred are
specified in detail in many cases, not just ad-hoc.

I wonder if the reason you are so bearish on the worldwide web is that it's
not a _clean_ distributed database.   It's quite messy.   But best is the
enemy of good enough, and the www is certainly that.

On Fri, May 13, 2016 at 3:52 PM, Dave Crocker <dcrocker(_at_)gmail(_dot_)com> 
wrote:

On 5/13/2016 11:20 AM, Ted Lemon wrote:

On Fri, May 13, 2016 at 1:56 PM, Dave Crocker <dcrocker(_at_)gmail(_dot_)com
    To date, we really only have two services that demonstrate open (ie,
    multi-administration) interoperability at Internet scale:  email and
    DNS.


TCP, IP, UDP, NFS, DHCP, TLS, SSH, FTP, HTTP, NNTP, SIP, ...



So, yes, I did intend my comment to be provocative and did suspect I was
missing one or another service.  But I also messed up, by not making clear
I was targeting open, /distributed/, /applications-level/ services. (And
yes, for this kind of discussion, DNS is an application.)

That is, I meant the qualifying test to be that there often is casual
interoperability across a /sequence/ of independent administrations, and
use by a very large fraction of the Internet.

Alia's BGP reference was the biggest surprise -- thank you, Alia! --
because I think it /does/ qualify and it hadn't occurred to me.

The problem with all of the others cited above is that they aren't at used
at scale or aren't really used with open, multi-hop interoperability.  Much
of the list, above, is for lower-layer protocols.

FTP and HTTP are simple, single-hop client/server mechanisms.  The latter
is, of course, widely used, but it's a one-hop service.  (In reality, of
course, the web has all sorts of additional hops, if one looks at content
distribution, and other mechanisms, but they are behind the scenes and
under tight control.)

NNTP is a very nicely distributed service, but it is not used at scale.

As I understand SIP use, the multi-hop mechanisms are another example of
tightly-control operational prior arrangement, behind the scenes.  So it
might qualify for "at scale" (though it might not) but it's operation isn't
sufficiently open -- ie, permitting /casual/ interoperation.

I'm a fan of xmpp/jabber, but it, too, simply hasn't attained sufficient
'at scale' use.

Hence my slightly-modified claim that, other than email and DNS (and, yes,
BGP), we have been strikingly unsuccessful at deploying new distributed
application services and getting them to be successful at scale.




d/

--

  Dave Crocker
  Brandenburg InternetWorking
  bbiw.net

<Prev in Thread] Current Thread [Next in Thread>