Three more;
* Does the NOTIFY message include the URI of the resource that has
changed (didn't see it, but I may be missing something). It should be
added if not; otherwise, you're baking in a one-to-one relationship
between monitoring channels and http URIs. There will be use cases
where the receiver of the NOTIFY doesn't have enough context to
associate the message body you're sending with a URI.
* The simplest and most efficient notification message to send is
"this resource has changed"; the client can then decide whether or not
to fetch an update, and combining any old response with a new one is
simple. It also assures that the subscribing client isn't required to
implement a cache. I would very much encourage you to support this
mode of interaction.
* Conversely, if you are you intent on allowing transfer of an entity
with a notification, you may want to consider allowing the transfer of
multiple entities, to allow the use of content negotiation (e.g., both
an English and French representation under the same URI, or a png and
a gif). application/http would support this.
Cheers,
On 11/12/2008, at 4:48 PM, Mark Nottingham wrote:
Sorry for the delay, finally off the road and had a chance to read.
Overall I like this document a lot, and I think it should be easy to
align its use of the 'monitor' link relation with the Atom-over-HTTP
protocol described in draft-nottingham-http-cache-channels.
A couple of comments / concerns:
* "...it MUST return a Link: header..." seems like it's too strong;
servers may want to be selective about when they offer monitoring,
and there may be other ways to advertise it that are more appropriate.
* Following on that thought, the idea behind the most recent Link
draft is to talk about link relations generically, and make Link
headers one use of them. Another use of them is in draft-nottingham-
site-meta (which is still under pretty active revision); this would
allow you to advertise a monitoring endpoint for your whole site,
for example.
* "It MAY return both." is ambiguous; think you mean SIP + SIPS.
* 3.4 Subscription Duration -- just a thought, is it worth tying the
default subscription duration to the freshness lifetime of the HTTP
response (as defined by HTTP)?
* 3.5.1 message/httpfrag -- did you consider using application/http
(defined in RFC2616)? I mention this because HTTP already defines
how to update a cache entry's headers based upon a HEAD or 304
response; it would be nice to reuse that if possible. Also, what
does making the status-line optional bring? The semantics of the
message would be clearer if it were required...
* 3.5.2 - Why require an ETag or Content-MD5? Why recommend a Last-
Modified?
* 3.10 - limiting notifications to one a second makes sense also
because this is the granularity of time in HTTP.
Specific responses below;
On 22/11/2008, at 2:53 AM, Adam Roach wrote:
On 11/21/08 2:07 AM, Theo Zourzouvillys wrote:
On Fri, Nov 21, 2008 at 4:12 AM, Adam Roach <[EMAIL PROTECTED]>
wrote:
Very interesting proposals. Admittedly, for the use cases SIP is
most
interested in, the results are not typically HTML, but the idea of
communicating these in the body may have merit. I'm curious about
which use
cases you have in mind -- do you imagine that there will be
circumstances
under which the HTTP server will have the ability to notify other
nodes of
changes, but not have the ability to include a Link: header?
absolutely - perhaps a resource hosted on a CDN, where you commonly
can not send custom headers.
You certainly can, in a cache-driven CDN (e.g., Akamai); the cases
where you can't aren't really CDNs as the industry understands them,
they're Web hosting farms (because they use replication; e.g.,
Amazon's new service, if I read it correctly). In that case,
providing a channel without coordination with them can be harmful,
because you can get into race conditions where you send the
notification before the content is fully replicated.
More to the point, if you're using a CDN (of either type), the most
likely value-add scenarios are using a monitoring channel to notify
the CDN itself of changes, and also the CDN providing a monitoring
server as a service.
or a blog, which may be hosted on some other platform (for example,
blogger.com) where you can edit the HTML but not the headers.
I'm getting somewhat exasperated at how often this argument comes up
in specification discussions; to date mostly at the W3C. If we
follow it to its conclusion, the only Internet protocols that we can
develop are those that work with free hosting sites that only
provide FTP access. Surely the Internet and the Web shouldn't be
held hostage to them.
That said, I agree that the draft shouldn't preclude other discovery
mechanisms, as mentioned above.
the Link header as defined in draft-notthingham-http-link-header
itself is only a way of representing the <link> header in HTML and
<atom:link/> header in atom entries outside of the content where
it is
not possible because the content-type does not have support. So it
already "just works" by creating a link registry entry, however
would
be good to explicitly point out that it could be anywhere a link
relation is allowed.
That's interesting -- Mark, do you have any input on the topic?
I'm having trouble parsing the above -- is the concern that some
people might believe that any link in the headers *has* to be
mirrored in the content? If so, that's the first time I've heard
this brought up, and it certainly isn't the intent. We can clarify
if necessary.
Cheers,
--
Mark Nottingham http://www.mnot.net/
--
Mark Nottingham http://www.mnot.net/
_______________________________________________
BLISS mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/bliss
