World: r3wp

memory resident db ?

Carl's blog needs a way to implement emoticons !

:(

Rebol is alive... again :)

Graham: good idea, I already wrote REBOL code to support emoticons 
in another project. I'll add it to the blog. (Note to myself : The 
link to Carl's blog source code disapeared in our blog template file, 
I have to put it back!)

Memory DB: even in memory if you count the overhead of a SQL layer 
+ storage layer, it would still be much slower than direct storing 
of data in a REBOL block! or hash!.

Doc: I know, but I also read the RubyOnRails book where they measured 
all this. A database scales better, and for low usage the performance 
hit you tahe is little. Files/memory is faster but collapses pretty 
soon, MySQL runs with constant overhead (and little enough, I might 
add). It is more a question of balancing performance with different 
load scenarios.

http://en.wikipedia.org/wiki/Comparison_of_web_servers
Cheyenne missing here :)

I was using rebol blocks to store greylisting data ...

However after a few hundred thousand greylisting entries .. it started 
to collapse a little.  I should have gone db ...

Doc, I am also one of your fans. I've been going to your web site 
regularly to see if Cheyenne was released. I was excited to be able 
to download it. But I have a problem. Cheyenne loads fine, but when 
I try to test the links on the index.html page they fail. What am 
I doing wrong?

Are you using the exe version or source version ? What OS ? Any crash.log 
file found ?

Both versions act the same. Win XP Home. No crash.log file found. 
 This is what loads:

[uniserve] Async Protocol FastCGI loaded
[uniserve] Starting task-master...
[uniserve] Starting HTTPd...

Can you give me details on what is failing precisely ?

Test CGI script

Test RSP script

Test RSP WebApp

What error message do you have ?

For the RSP WebApp test I get:

 The page cannot be displayed 

The page you are looking for is currently unavailable. The Web site 
might be experiencing technical difficulties, or you may need to 
adjust your browser settings.

For Text CGI script I get:


#!/usr/bin/rebol --cgi REBOL [ Title: "show" File: %show.r ] print 
"Content-type: text/html^/" print {Back

} print ["Script path :" system/script/path "

] print 
CGI Object :" print "

 foreach name next first system/options/cgi [ either :name = 'other-headers 
 [ print [
 name 
: 
] foreach [n v] list: system/options/cgi/:name [ print [
 n 
: " mold select list n ""] ] print "
 ][ print [
 name 
: " mold system/options/cgi/:name ""] ] ] print "
 if system/options/cgi/request-method = 

POST" [ vars: make object! decode-cgi make string! input if not empty? 
next first vars [ print " Variables passed :

 foreach name next first vars [ print [
 name 
: " mold vars/:name ""] ] print "
 ] ] print 

For Text RSP script I get:

Back

Timestamp: 


Request parameters :
HTTP Method: 
HTTP Port: 
Client IP address: 
Request headers :
Request variables :
Session :
SID: 
No session

Looks like the configuration parameters are not loaded...Did you 
modified your httpd.cfg ?

No.

However, since it wasn't working, I did set the includes to the SDK 
sources on my harddrive, and then reencapped.

For the tests, I get the same results whither Cheyenne is running 
or not.

Did you tried without encapping, just using the sourced version ?

The encapping procedure is special for Cheyenne, it needs to be run 
in normal mode first before encapping to preprocess the source and 
generate the .cache.efs file.

Btw you shouldn't get any result if Cheyenne is not running !

Are you testing using the http://localhosturl ?

OK, http://localhostworks. I have a lot to learn, as you can see. 
Thanks for helping me through this. I was just clicking on the index.html 
file in the www folder.

Can you make the session storage server a RebService, called from 
the various processes that need to read/write the data? That overhead 
might be less than the SQL overhead.

I haven't had a chance to review the Cheyenne source yet. Are the 
subprocesses called by a central broker? Perhaps that broker can 
manage the session data. That way any in-process accesses could be 
almost as fast as straight block access, with only API overhead.

Brian, that is exaclty how sessions are handled right now. Sessions 
are managed by central broker uniserve with service httpd.r and data 
exchanged with task-handlers subprocesses executing rsp

Well, if you package the session data in a manager API, you can export 
that API as a REBOL Service that subprocess can use as well.

are rebol services stable and fast?

Not sure. I remember that there was a version built on UniServe, 
same as Cheyenne, so we could use that.

Or was it built on BEER?

It's BEER I guess, not UniServe/

Does the REBOL/Plugin work with Cheyenne?

Doc, please see my posts in the Plugin-2 group.

Also, would be nice to have a link on your Cheyenne page to your 
Cheyenne blog.

REBOL/Plugin: yes, as any other HTTP client.

Blog: I'll add it this WE, I'm still doing adding some features to 
the blog, I'd like to test it a few days more with a low traffic.

I'd love if there were an LNS implementation built on UniServe. This 
would let you turn Cheyenne into a general purpose app server.

general? Rebol general,no? maybe XML-RPC would be interesting for 
non-rebol world, but it could be added too ....

I've been thinking of LNS syntax shims that would allow REBOL services 
to be called by non-REBOL clients.

Does Cheyenne have an option to limit which network adapters it will 
serve to? For example, I have 5 adapters on this computer (ethernet, 
wireless, 2 VMware, Hamachi), maybe 6 counting localhost, and I would 
like the web server to only be visible to the virtual adapters, not 
even opening a port on the physical ones.


I remember this being difficult in REBOL, that REBOL would always 
open server ports on all available adapters, so you would have to 
inspect the incoming connections and filter. Have you come up with 
a better workaround for this, or a way to do it properly?

IIRC via set-modes/get-modes, you can know what adapter is involved. 
I am afraid there is no other way how to prevent this. Opening a 
socket is OS thing imo, and I wonder if C socket functions allow 
you to choose your device?

They do, AFAIK. I do have the standard Windows firewall installed, 
but it would be better (lower overhead) to just not open the port 
on the wrong adapters in the first place.

It can be done at the C socket level, unfortunately, I don't know 
any way of doing that from REBOL.  I guess that's not supported in 
the current REBOL port!.

Can Cheynne support Ajax ?

Does each http request spawn a new instance ?