On 11/17/20 11:03 AM, Theodore Y. Ts'o wrote:
And while it is true that the http*protocol* has been evolving, it is *not* a "user interface"; it is a protocol which supports user interface programs,
To clarify, my point is that the provisioning and configuration of HTTP[S] services is driven by user interface considerations, since that's how people view "the web". (even though "the web" intends and always was intended to support more than just user interfaces). That's why, for instance, web content often breaks old URLs and the content authors/providers feel justified in doing so.
It's hard to make the same protocol suite serve multiple masters.
and as a protocol there has been quite good backwards compatibility, since people (especially in government) have insisted on running Windows 95 long past when it was safe and sane to do so (speaking as someone whose SF-86 was compromised as a result), and so web servers will support ancient http clients without any issues that I'm been made aware of.
What I've seen is that for http (not https) the backward compatibility in browsers has to date been very good, but the deprecation of older versions of TLS has broken lots of things. Anyway, that's what I was thinking of when I said "unstable". And now I learn that browsers are dropping support for http, which I think also supports the argument that the web protocol suite is unstable.
Keith
