[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Some cautionary comments
- To: "Mark Day" <firstname.lastname@example.org>
- Subject: Re: Some cautionary comments
- From: Fred Baker <email@example.com>
- Date: Tue, 16 Jan 2001 08:48:47 +0800
- Cc: <firstname.lastname@example.org>
- Delivery-date: Mon, 15 Jan 2001 17:20:31 -0800
- Envelope-to: email@example.com
Very much agree. I'll add my two cents...
At 01:33 PM 1/15/01 -0500, Mark Day wrote:
>1. Negotiation of parameters
>It is often tempting to defer a hard choice in a protocol design by saying
>that multiple possibilities will exist, with a negotiation at runtime. The
>unfortunate result is that every implementation has to deal with all the
>possibilities. In addition, negotiation structures complicate the process
>of interoperability testing.
Very much agree. IMHO, this was the downfall of the OSI protocol suite -
interoperability went pretty much out the window. It is in the working
group's interest to make the hard decisions, and to make them in the
direction of simplicity unless there is a very good reason not to.
>2. Hypothetically useful capabilities
>Since we are creating new protocols and enabling new kinds of system to be
>built, it is tempting to start imagining entirely new ways of using them.
The thing to do with these is pull them out into separate drafts, so they
can be explored by those interested in them, but keep them out of the base
specification. These become in some sense experimental procedures.
>This can be fun but it's not very functional.
It also tends to kill a lot of time, and the base protocol often develops
odd characteristics that are perhaps the right thing to do if the feature
is useful, but for most purposes is suboptimal.