Re: [apps-discuss] "X-" revisited

[Dirk Pranke <[e-mail]>]

On Mon, Jul 11, 2011 at 5:17 PM, Mark Nottingham <[e-mail]> wrote:
> On 12/07/2011, at 4:37 AM, Dirk Pranke wrote:
>
>> On Sun, Jul 10, 2011 at 6:18 PM, Mark Nottingham <[e-mail]> wrote:
>>>
>>> On 07/07/2011, at 1:30 PM, Dirk Pranke wrote:
>>>
>>>> On Wed, Jul 6, 2011 at 7:24 PM, Peter Saint-Andre <[e-mail]> wrote:
>>>>>
>>>>> Is there some kind of attack lurking here that we're not aware of?
>>>>> Parameter phishing or somesuch?
>>>>>
>>>>
>>>> No, that was not my concern. I am mostly trying to map your arguments
>>>> onto the way we're currently evolving the HTML APIs, which follow a
>>>> similar convention to X- (the vendor prefixes).
>>>
>>>
>>> I think the difference there is that the number of implementations is relatively small.
>>>
>>
>> Hmm. I agree that the number of browser implementations with
>> significant market share is fairly small, but I don't know that this
>> is unusual. I'd be hard pressed to think of a significant application
>> or transport-layer protocol that has had more than a half dozen major
>> (read: others have to adopt their extensions) implementations either.
>
> Right, but think of HTTP headers; any random person can (and does) add them, all of the time, and then they're in use.
>
> This is where "implementation" breaks down, because it's the "server" + frameworks + site-specific code. While there are a few handful* of HTTP server implementations, there are hundreds of frameworks, and millions of sites with their own code. Then there are conventions that people layer over top, for interop between sites / frameworks (e.g., Atompub).
>

I think I did not make myself clear. Just because some implementation
uses a header named 'X' does not mean that everyone else cares. We
ignore headers we don't recognize. We usually only care about header X
when a significant percentage of our traffic starts using it.

On the other hand, your point about there being a lot of frameworks is
valid, and I agree that there's a lot more than six of them :) Then
again, frameworks rarely create new headers in my (admittedly limited)
experience.

-- Dirk

> Cheers,
>
>
> * From what I can see, the number of HTTP server implementations is exploding as well, over the last few years.
>
>
> Mark Nottingham   http://www.mnot.net/
>
>
>
>