2.1.15 Web Replication and Caching (wrec)

NOTE: This charter is a snapshot of the 49th IETF Meeting in San Diego, California. It may now be out-of-date. Last Modified: 15-Nov-00


John Martin <jmartin@netapp.com>
Bill Maggs <bill_Maggs@palm.com>

Applications Area Director(s):

Ned Freed <ned.freed@innosoft.com>
Patrik Faltstrom <paf@cisco.com>

Applications Area Advisor:

Patrik Faltstrom <paf@cisco.com>

Mailing Lists:

General Discussion:wrec@cs.utk.edu
To Subscribe: wrec-request@cs.utk.edu
Archive: ftp://cs.utk.edu/pub/wrec

Description of Working Group:

The purpose of this working group is to define a co-ordinated caching and replication framework for the world wide web.

Replication of web services and caching of http responses are distinct paradigms that to a large degree require distinct solutions. However, it is also important that web caches and replicated services be able to coexist. Ultimately, the working group will define a framework for efficient, reliable, and predictable service in a web which includes both replicas and caches. Part of this framework will be a means by which replicas and caches may be automatically discovered and utilized by clients.

Because of the wide variety of caching and replication solutions which are already in use, the initial task of the group is to familiarize itself with the diversity of existing practice. The group will therefore produce a document which provides a taxonomy of existing caching / replication developments and terminology which would be used as a reference for later work. In parallel with this, existing protocols and standards which are not publicly documented will be submitted to the IESG for publication as informational RFCs by their respective proponents. Once the initial phase is complete, the group will make a recommendation as to whether further protocol work is necessary or desirable, and submit a revised charter for consideration by IESG and IAB.

Goals and Milestones:

Mar 99


First issue of I-D on taxonomy and terminology (Editor: I. Melve)

Mar 99


First issue of I-Ds on existing caching / replication protocols

Mar 99


Meet at Minneapolis IETF

Apr 99


First issue of I-D on design goals / research issues (Editor: J. Touch)

Apr 99


Final issues of I-Ds on existing caching / replication protocols

Jun 99


Final issue of I-D on taxonomy and terminology

Jun 99


Final issue of I-D on design goals / 688 research issues

Jul 99


Meet at Oslo IETF

Jul 99


Submit I-Ds to IESG for publication as Informational RFCs

Jul 99


Revise charter


No Request For Comments

Current Meeting Report

WREC minutes (IETF49, San Diego)
Tuesday 12 December 2000
Minutes taken by Ian Cooper (co-chair)
Co-chairs: Mark Nottingham
Ian Cooper

Presentation slides: http://www.mnot.net/papers/ietf-49-wrec.{ppt|pdf|htm}

The meeting was a combination of an update on the status of WREC as well as an introduction to possible future work within a new group, WEBI (Web Intermediaries).

Ian Cooper gave an update on the status of the WREC group and work. Ian and Mark Nottingham had replaced John Martin and Bill Maggs for the remaining life of the group, and gave thanks for their efforts in the past.

An update of the "Known Problems" document was given. The format of the document had been changed to move more "historical" interest items to the appendix. The revised document was almost ready for group last call.

The "Taxonomy" document had been returned by RFC-editor due to some concerns with the types and quantity of some references. Internet Drafts appeared to have been given as normative references (hopefully resolved by changed wording); the large number of URLs used in, and as, references was a concern. URL only references were removed where possible, and other URLs (e.g. for RFCs and Internet Drafts) were also removed. Document re-submitted to IESG and copied to the mailing list since it had not been re-submitted to the Internet Drafts queue.

Larry Masinter commented that the revised version needed to be submitted as an Internet Draft since there had been changes. Ian believed that since these were only editorial changes that this was an unnecessary step. [Later clarification indicates that the document needed to be re-submitted.]

In order to close WREC the outstanding work items (Known Problems and Taxonomy documents) needed to be finished or removed. Removal and a change to individual submissions was the chairs favored solutions in order to facilitate the move to new work.

Mark Nottingham described the motivation for new work in this field, within the context of a new working group. WREC identified issues in the Web infrastructure that need resolution; there is community interest in solving additional problems in the Web infrastructure. A presentation on two proposed areas of work was given: "Intermediary Discovery and Description" (IDD), and "Resource Update Protocol" (RUP).

IDD was proposed to resolve the known issues with interception proxies and the problems surrounding WPAD (reliance on PAC files; general lack of support).

Alex French commented that it may not be useful to put the intelligence into the client. The amount of information and processing may be too high for lightweight clients.

Michael Condry asked whether the locality of the discovery was dependant upon network or geographical topology, and that in response to Mark's belief that it should be dependant on network locality suggested that regional locality needs to be considered for services that are regional in nature.

Mark encouraged interested parties to subscribe to the WEBI mailing list. In response to a question to the floor, Ian clarified that subscriptions to WREC would not be automatically transferred to the WEBI list, and that the WEBI list would be announced on the WREC list.

Scott Brim asked whether there was a suggested date for having a stake in the ground on requirements. It was suggested that this should be Feb 2001.

Oskar Batuner commented that while IDD may be an academically correct thing to examine, this was an area where there was already a solution. There is a conflict of interests in the user choosing their intermediary vs. forcing users to connect through devices in the network. Most people (ed: network managers?) would still use transparent proxies (sic). The group should, perhaps, look at discovering interception proxies.

Mark Nottingham acknowledged that one could never get rid of all interception proxies.

Larry Masinter commented that the process is useful even if clients are not using it. Figure out the right thing to do then work out the transition process.

Mark Nottingham stated that IDD could be used for more than simple discovery of proxies.

John Martin noted that the two environments are not mutually exclusive. The proxy does not care how the request gets to it.

Lesley Daigle warned that there was a danger in suggesting transparency from the user and in automating meshes. There is a danger of getting involved in the relationship between the end user and the content provider.

With reference to a comment made in the presentation slides, Evan Baer commented that he liked the analogy between Web UA and mail UA (users think nothing of configuring their mail UA to use a local mail system, but this is not the case for Web access through a proxy). Perhaps there was scope for a new query type on top of DNS?

Mark Nottingham went on to give an outline for a "Resource Update Protocol" (RUP).

There were concerns on the use of XML for meta-data transfer.

With reference to the presentation slides, there was a query regarding URISpace. Mark was unwilling to say where this work was being done.

The proposed charter for WEBI did not propose a requirements document for RUP, but this was now seen as an omission.

Dan Li suggested that perhaps the protocol should be named WRUP (Web Resource Update Protocol).

Pei Cao queried whether invalidations and updates would be grouped together within the protocol. Invalidations can be small, with updates being much larger - this would force a design at the lower layer. Also commented that stronger authentication was necessary for updates than for invalidations.

Mark commented that authentication may cause many requirements.

??? suggested that SASL may provide a reasonable framework.

Phil ??? queried the use cases.
Mark: large number (hierarchy) of intermediaries where the state has to be transferred, without polling.
Phil: push technology?
Mark: perhaps

Ashish ??? commented on the one way channel, querying the restriction. It may be useful to have a back channel for monitoring. Mark commented that the nature of the channel was not set.

??? noted that there were common problems at the transport levels (multicast). Rather than reinvent that work, standardize interface at the edges, APIs. Dan Li also commented that the interfaces between transport and messaging mechanisms should be clearly defined.

Brad Cain queried which network layers were being considered for the work.

Paul Eastham stated that we must consider AAA issues (e.g. problems of accounting issues if things become cacheable). Some of the drafts (in other groups) already consider this. Needs further discussion.

Ted Hardie warned that the discussion was descending in many ratholes (in terms of proposing transport protocols at this early stage). Should consider the actual problems before considering the transport mechanisms.

Harald Alvestrand queried the scaling of the system, and suggested that we need to define how far the system scales. Mark responded that the system should be able to cope with 10,000 subscribers to the same object.

Michael Condry suggested that the charter needed revising.

Dan Li asked how the proposed work related to the CDNP distribution peering. Seems that there is a common signalling problem and how to package and group objects; concern that the work may be done twice. Mark Day commented that signalling may be inputs to design process. [ed: the proposed groups had already been requested to work together to identify common ground and what might be done where.]

Pei Cao asked whether the scope of RUP was for single or multiple administrative domains. Mark Nottingham replied that the work should be generic, not limited to either.

There was rough consensus that the two proposed work items were interested and well scoped. It was agreed that the charter should be revised to include reference to the RUP requirements document, and also to clarify the scope of the group in relation to other proposed groups in the area.

WEBI Mailing list details:
Subscriptions: webi-request@lists.equinix.com
To post to the list: webi@lists.equinix.com

Archive is currently within majordomo. Send message with body "index webi" for a list of files (or "help" for help with this feature). Web archive will be available shortly.