- From: <Dove@eworld.com>
- Date: Thu, 23 Mar 95 08:30:44 PST
- To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Dear Team, I represent a firm producing HTTP clients and servers for use in vertical mobile computing environments, using PDA-style devices. In the last year, we have produced a HTTP 0.9 compliant browser, and are currently collaborating with other firms to produce a HTTP 1.0 compliant browser and proxy server for wired or wireless access to devices which may or may not have an underlying TCP/IP stack. To this end, we have reviewd the draft IETF HTTP spec (0.1, dated 19 December 1994, expiring June 19, 1995) quite thoroughly, and it has been very useful in both the development of our implementation and in developing proxy servers for this and other applications. In our work, we have seen a fairly strong need to reduce the amount of document data transferred between proxy and mobile client - traditionally, this link is achieved via a slow-bandwidth or high-byte-cost connection (or more often, both) such as the ARDIS network or a 2400bd modem connection. As we approach incorporating HTTP 1.0 support, we have debated adding a subdocument query extension to HTTP, to support the following scenario: 1. The mobile client requests a document from the proxy, indicating it only wants a single "page" (of client-specified length). 2. The proxy fetches the full document (or perhaps only the first page, depending on the version of its source) 3. The proxy returns the desired "page" to the client. This page is an arbitrary range of bytes, which requires that the client have some mechanism of dealing with fragmented data (split HTML tokens, for example, or reuniting pieces of a split graphic as required). By allowing this "paging" to be performed by the proxy, which presumably has an inexpensive high-bandwidth connection, we can decrease transfer time and transfer cost. Our user studies have indicated that in a large number of situations, only a small segment of a document (usually, although not always, the beginning) is used by a user before pursuing another. Rather than using a start/stop scheme, we thought that this paging mechanism would be more robust. Based on our experience in mobile data transfer, we think an extension to the Request Header of HTTP 1.0 requests to permit the request of only a certain byte range may be of use to more than just us and our customers; as wireless access to the network grows, proxy servers used by wireless providers could offer this mode to lower transfer time and save customer costs, for example. Thus, we felt it would be best to call your attention to our efforts in the hopes that a unified standard for this mechanism could be made. We would thus respectfully submit an additional keyword to the Object header to denote the range of octets for a particular document being requested. This keyword would be known as Octet-Range, and perform as follows. BNF Description of Octet-Range in the Object Header: Octet-Range = "Octet-Range" ":" start "-" end start = 1 * DIGIT end = 1 * DIGIT Requests including the Octet-Range Object Header would generate responses containing only those octets in the Object Body offset by the start value (counting from zero) and through the end, or the end of the Object Body, whichever resulted in the smallest Object Body. It is an error to return Octets before the indicated start offset, or after the indicated end offset. A response incorporating the Octet-Range would return the Octet-Range of the request, and the associated octets of the Object Body as the Object-Body. In the event that the end offset was greater than the total number of octets, the Octet-Range Header still has its original value, but the Content-Length Object Header shall indicate the number of octets actually returned. (The client may be able to then deduce the true total length of the document as its start plus the returned number of octets without the need for a HEAD request). Your consideration of this matter is appreciated. Please feel free to contact us with any comments, questions, or feedback you may have. Best Regards, Ray Rischpater Software Craftsman AllPen Software, Inc. (408) 399-8800 voice (408) 399-4395 fax dove@allpen.com dove@eWorld.com
Received on Thursday, 23 March 1995 08:41:37 UTC