- From: Lisa Dusseault <lisa@xythos.com>
- Date: Tue, 17 Jun 2003 16:34:32 -0700
- To: <shadgar@cs.bris.ac.uk>, "'WebDAV'" <w3c-dist-auth@w3.org>
> I haven't expressed my question correctly - sorry. By WebDAV > and HTTP, in fact I meant comparing the process of > communication between the WebDAV servers and clients compare > to the HTTP ones. > > I agree with you that WebDAV is an extension on HTTP and > provide all functionality of the HTTP, added some new > functionality regarding the metadata, access control, locking > and so on. However, since there is more processing for those > new functionality in environment based on the WebDAV > protocol, I am wondering how much this would effect on the > speed. Also many functionality of the WebDAV can be > implemented by HTTP POST method, Soap and so on ( as you > mentioned like Microsoft FrontPage). So if the security is > the same and if the speed may be low (not sure), why would > the user choose WebDAV rather HTTP? > A WebDAV server can be implemented that has the same performance as an HTTP server, for the HTTP functions. THat is to say, as long as the server is being used only for HTTP functions, it would have the same performance and scalability. A thousand GET requests can theoretically be handled in the same time whether or not you've got WebDAV. When the server's usage includes WebDAV requests as well, obviously this can change things -- and now it really depends on the scenario. - When users start using WebDAV functionality, are they doing *more* requests? On top of the thousand GET requests per minute, are users now also doing a thousand PROPFIND requests? Obviously this is twice as many requests and takes more processing time on the server. - On the other hand, when users start using WebDAV does this *replace* HTTP requests? Imagine 500 of our 1000 GET requests are directory listings, so when WebDAV replaces the server generated directory listings, there are then 500 GET requests and 500 PROPFIND requests in an average period. Now the server can be *faster* because it should be faster to generate and send 500 PROPFIND responses than to compile 500 dynamically-generated GET responses with the same directory listings. That's because in the HTTP scenario the server is responsible for generating the presentation (in HTML) of the info, whereas in WebDAV the client is responsible for presentation. This kind of analysis can be extended. In general, and on average, I think we find that when the server has to do presentation it has more work to do, and thus has worse performance and worse scalability. WebDAV generally allows the server to do less presentation work, which should improve its performance and scalability (as long as the overall usage patterns do not change, which they might). An example of this was documented at Microsoft in the development of Outlook Web Access for Exchange 2000. When the server had to do all the presentation work itself, the OWA server could not handle so many users. When the server offloaded the presentation work using scripting, stylesheets and WebDAV, the number of users that could be handled increased by a very significant amount. If the client supported WebDAV without needing scripting and stylesheets (if it had all its own presentation logic) then even more clients could have been handled. So in a general authoring scenario, I would expect WebDAV to be much more scalable and high-performing than a HTTP interface that uses forms (like Yahoo PageWizards). Now let's consider FrontPage. FrontPage is very much like WebDAV in the semantic operations it defines. Thus, we would expect its performance to be very much like WebDAV. They both extend HTTP, it's just that FrontPage does it in different ways. But both FrontPage and WebDAV leave the presetnation layer to the client so they have similar characteristics. SOAP is an entirely theoretical case. There doesn't exist a set of authoring operations defined for SOAP. So how could we compare them? SOAP is a protocol for delivering XML payload but it doesn't define any schemas for payloads that would allow authoring operations, like LOCK and UNLOCK and metadata operations. If you were to write a new schema for SOAP to define authoring operations, since SOAP is based on HTTP as well it might be as scalable as WebDAV, but it would likely be less scalable (because there is overhead in SOAP that WebDAV doesn't seem to need). There are a number of factors you didn't mention, beyond performance/scalability, why a user would choose WebDAV authoring rather than a non-standard HTTP-based client authoring solution. - WebDAV allows clients to syncrhonize data in order to maintain sites (sitecopy), for offline usage or to run automated backups (Xythos WFC). - WebDAV allows clients to control the UI. Your UI can be Command-line (like cadaver) Mac-like (WebDAV-FS, Goliath) Windows Explorer (Xythos WFC, Web Folders) A search-oriented file "locator" rather than explorer (KCura) - WebDAV allows users to save straight from Office and Adobe authoring applications (and any other app that chooses to support WebDAV). This saves the upload step. - WebDAV allows the user/client to choose a language and supports sight/hearing disabled usage better than HTML. - WebDAV allows locks to be obtained automatically by the authoring application. HTTP-based authoring sites either don't support locking or make the user get locks manually (like www.sharemation.com) Beyond this I suggest you preorder my book, though it's not out quite yet. (http://www.amazon.com/exec/obidos/tg/detail/-/0130652083/qid=1055892821/sr= 1-1/ref=sr_1_1/103-3410589-6207009?v=glance&s=books) Lisa
Received on Tuesday, 17 June 2003 19:34:34 UTC