World: r3wp

I have also a smtp protocol that I will use to send mail.  Why not 
use the existing smtp that comes with Rebol?  My one uses Vincent's 
dig to find the email's mail server, and sends it directly.

This bypasses all those problems that some people have with authenticated 
smtp servers.

From the documentation, "Uniserve's purpose is to offer a simple 
but powerful solution for programming client and servers applications 
that can be combined with View interfaces easily."


The uses for the Asynch that I can come up with are: 1. Backup after 
transactions. 2. Journalizing / reporting after a committed transaction 
update 3. Database access 4. Queuing of side-effect events that are 
not needed for the response in the user interface. 


Aside from that, my understanding is that an synch event could be 
made asynch to allow a lower end server to process the workload.

Does anyone have any other uses from the 4 noted?


Does anyone have a simple sample of tying Uniserve into the View 
interfaces?


Are there any results of driving some load using uniserve versus 
synchronous or other solutions like threaded models?




I am trying to get my head around this a bit more and want to rely 
less on the Doc-likes-it-some-it-must-be-the-right-thing-to-do approach 
that I have used before.


I read the Medusa documentation that Doc refers to but it did not 
lead me to any good application examples.

My replication server/client is using View and Uniserve

Thanks Graham.  Your response came in while I was composing this.... 
OK I have a simple View interface displaying a uniserve response 
so I follow that.  With a view user interface I can see you could 
get the data needed for multiple tab pages asynchronously and have 
some saving there.   For a script serving out html pages, it could 
be broken up so that asynch go after independent parts but they would 
have to be all emitted to the browser at once.   Is anyone doing 
that with Uniserv or other?

don't understand the scenario .. are u talking about writing a server 
or client application?

In that ramble, I had both - I have a View client app that I want 
to know how to access Uniservices to make the programming easy and 
I also want to know how to Uniserve if I want to run  a CGI that 
connects to a database, does some credit application calculations, 
and emits html to the client.   For this CGI script, I would get 
both performance improvements and an easier programming model ... 
if I am reading the doc correctly.

Didier is using Uniserve on client side for his forum program - enables 
async downloading of new messages.  Look for ( I think ) lecture-forum.r

My webmail using Uniserve's http server (Cheyenne) does database 
connections to retrieve mail http://www.compkarori.co.nz:8001

I have a single odbc connection that is re-used by all the clients 
so that I do not have the overhead of opening/closing a database 
connection  for each transaction.

Thanks Graham. One like a background refresh of more recently posted 
messages would be a great example.   I looked at Didier's code for 
lecture-forum.   If it uses Uniserve in the version that I located, 
it is very subtle.  I'll ask him if he can illuminate it for me.

Open/skip will be very helpful to allow file resume in Cheyenne.

you mean open/seek, no? :-)

yeah ...

Any simple examples on using "stop-at" with uniserve to pick up posted 
data??? Can only dig up "GET" ??

stop-at binary seems to work, but not stop-at string!

Ok, I got it.. seems you need to have the "stop-at" in both the on-new-client, 
AND on-received

I have another question though.. with the httpd service, where is 
the actual data?  I get the headers, and i get the length of the 
data, but not the actual binary.. im submitting a form using POST... 
GET is fine.

Does lfred know ?

Not unless you taught him.

when i submit a basic form.. the only binary i receive is "POST / 
HTTP/1.1".. none of the other values?

[thru "Content-length: " copy post-bytes to crlf | none]

so, it must be post-bytes :)

is the number of bytes posted.

and request I guess holds the post data itself.

I think the problems was trying to toy with the URL, rather than 
processing with a .cgi script.

the httpd is trying to serve up a page, and i wouldn't let it.??

One can get all fancy playing with ports and protocols, but in the 
end, there's alot to be said for DBI

(CGI rather :)

Ahh.. your left hand was one row down and shifted to the right

That, or my brain was thinking of doughnuts.

no, it was twisted clockwise!

Ok, the problem with POST data was with domain forwarding, and not 
with Uniserve.. GET works fine, POST didn't.

Quick questions about Uniserve:

1) Can you encap it
2) Is the http daemon reliable
3) Does the http handler support CGI/Reblets
4) Can the http daemon listen on multiple ports

Encap: You have encapping with external scripts in detective? should 
be a selfmade do then, to check authoirity. But would extract some 
files when running.

cgi: i have read yes, launches external script. i guess that part 
needs little change, encap gets scriptname as arg.

reliable - ask graham. And its donckimbel and he sounds satisfied 
with it.

multiple ports - i guess. it can run multiple protocols, as plugins. 
so run the same plugin a few times on different ports.

5) DOc has an improved version which some people got. The webpage 
sounded like he is close to replacing apache.

1 -3 yes.  Haven't tried 4.  Think it should be able to though.

1) Yes, by the use of an encap-fs script and some little changes 
in your script.

Hello, you could find a simple bench study I've made for my job about 
Uniserve vs Apache, on the Rebol Documentation Project (see http://www.rebdocproj.org/article.php3?id_article=181). 
It's in french, but with some charts.  Uniserve is very close to 
Apache 1.3 (and 2.0, not published).

Uniserve under 1.3  when trying to run the CGI example on the main 
http page. .. ## Error in [uniserve] : Async Write phase failed !

Nenad, do you have any docs on encapping Uniserve?

No, but next release includes an 'encap-fs library and an example 
of encapping Uniserve with all dependencies and modules.

How's production coming?

Next release will be out when I'll find time to package it. It also 
needs new documentation to better explain the concept behind the 
framework.

Doc - so - will you release mysql protocol as 1.0 stable?

I think that we should finally go 1.0 after all those years - your 
code is of proven quality ...

I need to first fix the "bad handshake" random issue.