On Jan 17, 8:10 pm, Eran Hammer-Lahav <[email protected]> wrote:
> On Jan 17, 6:09 am, Nils <[email protected]> wrote:
>
> > In my opinion Metalinks are a representation of the actual resource
> > BECAUSE they are a description of something that actually exists, and
> > are not merely a link/pointer/reference.
> > Hence TCN is correctly used here.
> > And hence there is no problem at all.
>
> This is a bit of a  contradiction. Something cannot be a
> representation of a resource and a description of it at the same time.
> But of course people can (and have) stretched the meaning of
> 'representation' to be, well, pretty much anything.

A metalink is a representation of a resource, like a JPEG is a
representation.
metalinks may be far more "degraded" and contain pointers to better
version.
A description is always also a representation of the resource, it
usually is just (heavily) degraded.
(My philosophical view)

> > And I think a philosophical discussion of  "representation vs.
> > description vs. complete view" in general and of the intension the
> > authors of 1997 and 1998 RFCs had  in particular is non-sense.
>
> I am not sure how 1997/8 are related to this.

I meant the years HTTP/1.1  and TCN where published receptively. Sorry
for the misleading wording.
>
> > There is a valid point for Link:, e.g. for referencing previous
> > chapters of a document or telling the client "hey, I have an
> > alternative representation here you didn't think of requesting
> > yourself you might still be interested in", i.e when you actually want
> > to reference/point another resource.
>
> Link defines a relationship between any two resources. "My Metalink
> description" is a valid relationship.
>
I agree. And it might be a good thing to have Link pointers for dumb
clients.
But Link doesn't fulfill the propose/use case of TCN as used by
metalink.
Futhermore it adds more load to the server as it has to (start to)
serve the download in question.

> > When it comes to metalink TCN, the client explicitly has to ask the
> > server for this representation, and hence there is no problem for non-
> > metalink clients receiving the fairly degraded (to them at least)
> > metalink view.
> > (Maybe this should be clarified in the Metalink spec, especially that
> > servers should assign a pretty low q=3D-value to metalink
> > representations of a resource to avoid sending out metalinks to non-
> > metalink clients).
>
> Again, there is nothing 'technically' wrong with this approach any
> others have taken a similar position that a descriptor can be
> collapsed into a single resource URI.
The TAG members argue that our doing basically violates the indent of
the RFCs in question, which makes it technically wrong, no?

> But a consensus is building that
> this is the wrong way of doing things.
Unless there is a right way (and like mentioned above Link doesn't
fulfill the requirements (to reduce server load, especially number of
requests)) such a consensus is useless.
For avoiding any given "wrong way" there has to be a right way to
start with.

> What you might want to
> consider, since you find the semantic discussion as nonsense (which I
> can respect) is the deployment ramifications of using the Accept
> header. Many platforms limit access to such headers.
Such platforms likely limit everything else, too, like a potential
Link header.

> some proxies
> mishandle Vary headers (which BTW, the spec should require with to any
> Accept reply).
So should we now abandon TCN altogether, even for the "blessed" use
cases?

> and some providers will not allow using it on their
> servers.
But there is a guarantee you may use Link headers?

> You might want to read John Panzer's view of this [1].
The only "new" discussion point would be "Extendable - limited to a
single content-type for metadata, and does not allow any existing
schemas (with well known content-type)."
TCN Accept allows to specify multiple types and even associate
preference with those.
For example, DownThemAll! currently sends an Accept like this (Firefox
default + metalink):
Accept: text/html,application/xhtml+xml,application/xml;q=3D0.9,*/
*;q=3D0.8, application/metalink+xml;q=3D0.9
I could imagine other headers like the following would do a great job
too:
Accept: */*;q=3D0.8, application/metalink+xml,application/x-
bittorrent;q=3D0.9
Meaning: Return me a metalink or a torrent if you go some, else return
me what you got instead.

Other than this, he details some great ideas on how to overcome some
potential issues.

>
> You can also support both approaches, allowing providers to accept the
> Accept header, but also declare the availability of metalink using the
> Link header. Dumb clients can then do a HEAD on the URI if they cannot
> use Accpet.

I'm all in favor to adding Link to metalink, but not replacing TCN,
but instead additionally to it.

I also read your overview about different discovery methods [2].
Great overview.

TCN, however, still works best for the given use case, and except for
some "philosophical" issues and implementation/deployment obstacles I
don't see any show-stoppers there.

>
> EHL

Cheers
Nils

>
> [1]http://www.abstractioneer.org/2008/11/discovery-metadata-is-just-
data...

[2] http://www.hueniverse.com/hueniverse/2008/09/discovery-and-h.html

(Note: Resent after mistakenly sending only to Eran)
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"Metalink Discussion" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/metalink-discussion?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to