[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [wsrp] Client header propagation to user-agents
I think using the combination of namespacing rules and knowledge of this namespace whithin the portlet's client side code is of less use. For this case the portlet must be aware that it's running in a WSRP environment which is what we always want to prevent, right? This would break with different bridges. Mit freundlichen Gruessen / best regards, Richard Jacob ______________________________________________________ IBM Lab Boeblingen, Germany Dept. 2289, WebSphere Portal Server Development 1 WSRP Team Lead WSRP Architecture & Standardization Phone: ++49 7031 16-3469 - Fax: ++49 7031 16-4888 Email: mailto:firstname.lastname@example.org Subbu Allamaraju <email@example.com> To 02/13/07 09:11 PM Stefan Hepper <firstname.lastname@example.org>, wsrp <email@example.com> cc Subject Re: [wsrp] Client header propagation to user-agents Good points. From an implementation point of view, setting a companion cookie is an option. However, as you point out, there are browser limits on the number of cookies stored per domain (starting from IE's 20 to Safari's 1000+). Two possible solutions: a. Store the mapping information locally on the consumer b. Require some metadata from the producer on each cookie whether it needs to be proxied to the browser or not In general, storing mapping info in cookies may not be work out. When cookie limits are reached, browser will start dropping some cookies, and the consumer may lose the mapping information forever. Subbu Stefan Hepper wrote: > Hi Subbu, > with requiring that you re-write each cookie and that duplicate each > cookie on the client as you also require to remember somehow the > original domain, path, port. Given that the recommendation is to allow > storing 20 cookies per web server this would mean that only 10 cookies > per consumer are likely to be supported. Isn't that quite restrictive? > That would only be needed if you really want to access the cookies on > the client, so maybe it would make sense that the consumer indicates if > that is really the case or not and thus allow the consumer to make the > re-writing accordingly. Also the namespacing is only needed for the > client access use case. > > Stefan > > Subbu Allamaraju wrote: >> During last week's interfaces call, I was asked to outline a proposal >> to extend header propagation all the way to the user agent. This is >> in addition to the proposal submitted by Rich last week to transport >> such headers within the protocol. >> >> Here is a quick summary of the text below. >> >> The Set-Cookie/Set-Cookie2 headers can contain "name", "value", >> "comment", "domain", "max-age", "pat", secure", "version", "discard", >> and "port" attributes. >> >> The proposal is to require the Producer to use the namespacePrefix for >> generating the cookie name, and require the Consumer to rewrite the >> "domain", "path", and "port" parameters. >> >> (The "port" parameter is introduced RFC 2965 as part of the >> Set-Cookie2 header. This is not widely used.) >> >> The net benefits of this proposal are: >> >> - Portlets can write script to access state contained in cookies. This >> opens up the door to support more web applications over WSRP. >> >> - Consumers can use browsers as a cookie store instead of managing >> cookies locally. >> >> Please comment on before the next TC/Interfaces calls. >> >> Subbu >> >> ----------------------------------------------------------------------- >> >> Section 6.1.16 MimeResponse >> --------------------------- >> >> The following rules apply to cookies returned by the Producer via the >> clientHeaders structure with a name of Set-Cookie or Set-Cookie2. Such >> clientHeaders correspond to the Set-Cookie and Set-Cookie2 headers >> specified by RFC 2109 and RFC 2965, and the values of such headers >> MUST follow RFC 2109 or 2965 for all cookie parameters. >> >> The Producer MUST prefix the names of cookies with the namespacePrefix >> supplied by the Consumer. When a cookie includes any of domain, path, >> or port parameters, the Producer MUST ensure that their values >> correspond to any domain, path or port that it expects the Consumer to >> submit future requests to. Consumers may ignore cookies whose names >> and parameters do not follow these statements, as such cookies may >> conflict with other cookies managed by the Consumer. >> >> When a Consumer receives a clientHeaders element with a name of >> Set-Cookie or Set-Cookie2, it SHOULD return HTTP headers named >> Set-Cookie or Set-Cookie2 to the user-agent subjected to its >> security-policy restrictions in addition to those rules specified by >> RFCs 2109 and 2965. If the received cookie includes any of domain, >> path or port parameters, the Consumer MUST replace those with >> corresponding values for the Consumer such that those values remain >> valid for the user-agent receiving the cookies. The Consumer MUST use >> the Producer-supplied values for other parameters. >> >> If the names of clientHeaders received by a Consumer correspond to any >> caching response headers specified in RFC 2616, the Consumer MUST >> interpret those headers as guidance related to caching the markup or >> resource provided by the Producer. The Consumer MAY propagate such >> headers to the user-agent subjected to its own caching policies. >> >> Note that some Consumer implementations may prohibit transporting >> clientHeaders received during the markup generation phase to user >> agents. Portlets counting on user-agents receiving such headers must >> program themselves to account for this behavior. >> >> >> Section 6.1.10 ClientData >> ------------------------- >> >> When a Consumer receives Cookie or Cookie2 headers from the user-agent >> corresponding to cookies previously returned by the Producer via the >> clientHeaders structure, the Consumer SHOULD include clientHeaders >> named Cookie or Cookie2 in requests to the Producer subjected to its >> security-policy restrictions in addition to those rules specified by >> RFCs 2109 and 2965. If the request from the user-agent corresponds to >> serve a resource via HTTP, the Consumer SHOULD include corresponding >> HTTP headers named Cookie or Cookie2 subjected to the same policy. If >> the Consumer rewrote the domain, path, or port parameters, the >> Consumer MUST replace those values with those originally supplied by >> the Producer while generating the values. >> >> When a Consumer receives caching related request headers, it MAY >> propagate such headers to the Producer either via the clientHeaders >> structure (for those operations that return markup) or as HTTP request >> headers for resource generating requests over HTTP. >> >> >> >> _______________________________________________________________________ >> Notice: This email message, together with any attachments, may contain >> information of BEA Systems, Inc., its subsidiaries and affiliated >> entities, that may be confidential, proprietary, copyrighted and/or >> legally privileged, and is intended solely for the use of the individual >> or entity named in this message. If you are not the intended recipient, >> and have received this message in error, please immediately return this >> by email and then delete it. >> >> > > _______________________________________________________________________ Notice: This email message, together with any attachments, may contain information of BEA Systems, Inc., its subsidiaries and affiliated entities, that may be confidential, proprietary, copyrighted and/or legally privileged, and is intended solely for the use of the individual or entity named in this message. If you are not the intended recipient, and have received this message in error, please immediately return this by email and then delete it.