- From: by way of Henrik Frystyk Nielsen <frystyk@w3.org>
- Date: Mon, 13 Jul 1998 18:45:58 -0400
- To: www-http-ng-comments@w3.org
I've some basic comments to the general directions of this HTTP-NG: The HTTP-NG protocol isn't according the architectural model a Hypertext Transfer Protocol. How about HOPP: hyperobject processing protocol? Most of the things of the HTTP-NG protocol layer could done by CORBA and IIOP today. The garbage collector and the human readable URIs for type ids are nice ideas. But I'm not sure, if these ideas can rectify the huge work to be done for a new object based distributed architecture. For providing the old HTTP functions you must establish an object infrastructure over the HTTP-NG protocol layer. Between the logical hypertext objects you have to define some sort of protocol providing the Hypertext functionality. I regard this protocol as the real new HTTP in HTTP-NG. The SMUX idea of multiplexing TCP connections is a good idea, because filedescriptors of firewall proxies are a limited resource. But a lot of the issues here are solved with HTTP 1.1. If SMUX proves more generic and gives more performance than IP-tunneling or some other kind of TCP over TCP, it will be of use. I believe the whole thing is called HTTP-NG to solve some fund raising issues for a distributed architecture research project. HTTP-NG is not HTTP anymore, it solves a different set of problems. Personally I fear that HTTP-NG is losing the most important feature of HTTP 1.x: simplicity. The 12th fundamental truth about networking can be found in RFC1906 (http://info.internet.isi.edu:80/in-notes/rfc/files/rfc1925.txt): (12) In protocol design, perfection has been reached not when there is nothing left to add, but when there is nothing left to take away. There is a lot of truth in this april fool's day RFC in a way as in Dilbert cartoons. BTW I think this 12th fundamental truth about networking is derived from a sentence of french aircraft pilot and author Antoine de Saint Exupery. He describes the beauty of his aircraft in the same way. The title of the wonderful book is "Wind, Sand and Stars" (http://www.amazon.com/exec/obidos/ASIN/0156970902/002-5939310-9331441). I would like to add two more thruths to that list. Complexity is easy, simplicity is a hard to achieve, (From a text book about the formal specification language Z.) Standards are nice, beause there are such a lot of them. (From a book about client-server-computing.) You see, I like such one sentence wisdoms. Ciao, Uli Kunitz
Received on Monday, 13 July 1998 18:47:12 UTC