Hilarie et al,
I'd like to make a few comments off the top of my head. I understand that
the draft is just getting off the ground, so please don't take
these thoughts as negative criticism. It is probably the case that the most
useful discussion will take place at Pittsburgh, so think of this as me
getting some questions in early.
Overall, some background about both the project and the parties involved
would be most helpful, as well as an indication of the timeline intended.
I came away from the draft with the impression that it reads like a product
description in some parts. If your intention is to provide an extensible
framework, it would be more useful to break it down into an overall
architecture, and then identify mechanisms within that lend themselves to
standardisation. As it sits, the document's goals aren't particularly clear
or well-realized.
In particular, both APIs for 'proxylets' and 'ruleset matching' are
mentioned. Is the intent to standardize these, or are they merely concepts
that have been introduced?
Some of the terms used in the document don't agree with established
nomenclature, or seem product-derived. In particular, 'proxy' and
'surrogate' are misused throughout. Tighter integration with the WREC
taxonomy would be advisable.
You also mention describing a framework for multiple protocols. I wonder at
the practicality of this, as not all protocols are message-based. Doing so
increases the scope of the project outside of WREC, and is quite ambitious.
The more I work on defining the surrogate role, the more it becomes
apparent that there are several, separate areas needing consideration, not
just specification of a role. This architecture seems to touch on many of
them, and I think there should be some interesting discussion as to what the
appropriate scope for it is.
To me, the tasks in this space are (just an overview, I'll be happy to go
into detail if you'd like):
* Defining new roles
including surrogate, 'proxylet', 'remote callout server'
* Describe how they interact with current standards, including the HTTP
I particularly liked your section 7.1, AAA, and the relationships it
illustrates. However, I'd take this further; they beg questions of
application of cacheability, error behaviours, and assignment of metadata
for each target.
One central question is that of entity identity, which Jeff Mogul brought
up in Lisbon. A good portion of your draft is devoted to describing
mechanisms that change entities 'in flight', and this concern needs to be
addressed.
Also, some configurations will give rise to 'synthetic objects', which, do
to URI rewriting or local satisfaction, do not exist. The ramifications of
this practice need to be examined.
* Describe how they interact with other efforts
The obvious item here is ICAP, but I'm also curious to see how this type
of architecture interacts with things like P3P, WebDAV, etc.
* Describe preferred methods for enhancing functionality
In the surrogate case, the largest concern seems to be making the device
authoritative for the objects it serves, and the problems this introduces.
* Establish reusable mechanisms
This can include such things as configuration, administration,
security/access control, application of metadata.
I look forward to talking at Pittsburgh, and seeing if there's some common
ground in our interests.
Cheers,
On Wed, Jul 26, 2000 at 08:34:57PM -0600, Hilarie Orman wrote:
> The Extensible Proxy Services Framework has a new website,
> www.extproxy.org
> A mailing list, subscribable at ietf-openproxy-request@imc.org
> and a workshop announcement:
>
> Call for Participation: Workshop on
> Sophisticated Content Services for the Network:
> Extensible Proxies
>
> September 13, 2000
> San Jose Convention Center
> San Jose, California
> http://www.extproxy.org
>
> This workshop will help to define a new platform for the Internet, one that is essential for engineering the delivery of sophisticated content forms and is also essential for bringing a new level of sophistication to content delivery. We are seeking participants who can bring to bear architectural innovations, new services ideas, and to either inspire or be inspired by a new direction in the evolution of the Internet.
>
> In today's Internet, the infrastructure creating the "the net" is pointing towards a new network service model. The Web is becoming clients or user agents that consume information, content servers that provide the basic information and units of Web infrastructure which are the Proxies (or Proxy-Cache). These critical infrastructure devices provide the opportunity for enhanced services that expand the value of content on the Internet. Examples of such services are:
>
> · content adjustment - adjusting and/or assembling the content ultimately being delivered to the user. Today we see this in content assembly for advertising.
> · presentation adjustment - modifying the content presentation to be more suitable for the client. Today we have WAP and other gateways to adjust content.
> · semantic management - evaluate content against classification criteria such as virus detection, topic, legality, etc.
>
> Extending the notion of content services leads us to operations such as:
>
> · Data dependent cache functioning - cache management strategy is a function of the nature of the cached data. Content providers are already acting on the requirements for caching streaming multimedia data.
> · Web accounting - tracking Web usage independently of the content server or user client. Multiple network sharing - objects obtained from a data network can, for example, feed a broadcast system.
>
> A recent Internet-Draft (http://www.ietf.org/internet-drafts/draft-tomlinson-epsfw-00.txt) proposed an architectural framework for standardizing the components and APIs that can be used to build these services. This is the Extensible Proxy Services Framework. We are holding a working to present and discuss concepts needed for standardizing components in the Proxy system.
>
> Suggested areas for papers or panels are
> High-performance architecture requirements and engineering criteria
> Innovative services on proxy platforms (e.g. messaging, geographic positioning, etc.)
> Languages for extensible services
> Dynamic service extensions: delivery, loading, etc.
> New communication protocols
> Other new platform architectures for Internet infrastructure
>
> Papers and panels must be submitted by August 31, 2000. Please send a title and abstract for papers to condry@eng.sun.com; panel proposals should include title, abstract, and names of committed panelists and should be sent to horman@novell.com. Responses will be sent by September 3, 2000.
>
> Registration information will be available at http://www.extproxy.org
-- Mark Nottingham, Research Scientist Akamai Technologies (San Mateo, CA)
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:29 MST