[V4RB] Writing a simple server

[John DeSoi]

>At 10:09 AM -0500 2/11/03, jda wrote:
>>
>>There are some reports of the RB4 socket failing when multiple
>>connection requests arrive at the same time, and because of this I'm
>>destroying and re-creating my socket every other hour, and so far it
>>runs pretty flawless.
>
>I think that the RB 5 SuperSocket class is going to handle this 
>automatically -- another reason to wait for that to deploy.

At the lower level TCP layers, a web server opens multiple listening 
streams on the specified port. This is the only way not to miss a 
request. Some browsers open multiple TCP streams to the same server 
to handle a request that has mixed content - e.g. text and images. If 
you are not listening on multiple streams, you are going to get 
errors on the browser side. I don't know anything about RB, but it 
sounds like the current socket class just opens one listening stream. 
Perhaps the new version will hide these low level details.


>What is proper Authentication (and how do you do it)?

There are several ways to implement authentication. The two most 
common are "Basic" and "Digest". Take a look at the HTTP specs for 
pointers.


>
>Oh, and BTW, is there a maximum number of characters you can send in 
>a reply (like 32767)? That was a limit cor cgi's in OS 9, and I read 
>somewhere that it was imposed by AppleEvents, not by HTTP.

Right, there is no limit imposed by HTTP. HTTP 1.1 required that you 
put the content length in the header when you send back the response. 
Using the older HTTP 1.0 protocol (still supported by all browsers as 
far as I know) you can send as much as you want and then close the 
connection to indicate the end of the response.

Best,

John DeSoi, Ph.D.