[Potluck] a lightweight web framework

[Potluck] a lightweight web framework

[Nala Ginrut]

Sorry for the late guys!
I planed to give you a lightweight web-framework, and you may use it to
build your own site with Guile easily. Besides, with Guile's inner
web-server, you could have 10K concurrency performance.

But life is no so perfect, I was sick for few days, so I'm late for the
potluck. When I tried to polish it more beautiful, I found a bug in web
module (anyway it's a nice discover for a hacker). 
Seems that I have to release it as a half-baked one :-(
But good news is that I'll continue to polish it.

I put here:
https://gitorious.org/glow/artanis

Let me introduce it:
* very lightweight: the core artanis.scm almost 300 lines, easy to hack
and learn for newbies.
* a relative complete web-server implementation, include error page
throw and all the HTTP method(you have to specify your own handler)
* 10K concurrent performance for the server, takes advantage of the
Guile inner server. IIRC, andy once said it's 9K+ for static pages.
It's enough for you own site/blog.
* sinatra like style route, that's why it names "artanis" ;-)
* Database support(now use guile-dbi), mysql/sqlite/postgresql. But it's
easy to port to other database binding. (but I like dbi)
* session support (thanks for andy's advice)
* HTML template of SXML (very easy to use for Lisper)

I have no time to write a tutorial for it before I release, but there'll
be one soon. And you may read test.scm/blog.scm to get to know how it
works. 
https://gitorious.org/glow/artanis/blobs/master/test.scm
https://gitorious.org/glow/artanis/blobs/master/blog.scm
(for blog.scm, you may need blogdb.mysql script to build your database
quickly)

You may write a blog in ten minutes if you're familiar with web develop
and SQL.

PS: and I have to mention that bug, I believe it's a bug.

When the server-handler get the request, I found the uri in request have
no 'host', it's #f. It causes trouble for me to implement url redirect
mechanism, which used to implement admin authentication. 
I do think uri should keep 'host' value because it's useful for later.
And it's OK for 'read-request-line', it'll parse and store 'host'
correctly.

I think there's some link in the inner server module, which dropped
'host' value or created a new uri and throw the old-correct one. 
Any comments?


Sorry again for the half-baked work, it looks no cool. But I've ever
planed a perfect one... :-(

Re: [Potluck] a lightweight web framework

[Daniel Hartwig]

On 17 February 2013 13:03, Nala Ginrut <nalaginrut@gmail.com> wrote:
> PS: and I have to mention that bug, I believe it's a bug.
>
> When the server-handler get the request, I found the uri in request have
> no 'host', it's #f. It causes trouble for me to implement url redirect
> mechanism, which used to implement admin authentication.
> I do think uri should keep 'host' value because it's useful for later.
> And it's OK for 'read-request-line', it'll parse and store 'host'
> correctly.

Most HTTP requests will *not* include an absolute URI.  Instead, the
request line contains only the path.  This is not a bug.

There is a header, host, that can be used to fill in the blank /if it
is present/.  Doing this automatically in the web module is too
prescriptive; instead, each server should do this for itself as it
deems appropriate.  Also, I would just inspect the host header
directly and *never* manipulate the Request-URI.

<http://tools.ietf.org/html/rfc2616#section-5.1.2>

The host header is typically used by a reverse proxy or virtual host
provider to dispatch to the appropriate site handler/module, which
(usually) should not care what its hostname is.

>
> I think there's some link in the inner server module, which dropped
> 'host' value or created a new uri and throw the old-correct one.
> Any comments?

What makes you think that?

>
>
> Sorry again for the half-baked work, it looks no cool. But I've ever
> planed a perfect one... :-(
>

Lets see it get finished then :-)

Regards

Re: [Potluck] a lightweight web framework

[Nala Ginrut]

hi Daniel!
First, I must appreciate for your encourage! That makes me happier
though I'm still weak and headache. ;-)

On Sun, 2013-02-17 at 14:14 +0800, Daniel Hartwig wrote:
> On 17 February 2013 13:03, Nala Ginrut <nalaginrut@gmail.com> wrote:
> > PS: and I have to mention that bug, I believe it's a bug.
> >
> > When the server-handler get the request, I found the uri in request have
> > no 'host', it's #f. It causes trouble for me to implement url redirect
> > mechanism, which used to implement admin authentication.
> > I do think uri should keep 'host' value because it's useful for later.
> > And it's OK for 'read-request-line', it'll parse and store 'host'
> > correctly.
> 
> Most HTTP requests will *not* include an absolute URI.  Instead, the
> request line contains only the path.  This is not a bug.
> 

> There is a header, host, that can be used to fill in the blank /if it
> is present/.  Doing this automatically in the web module is too
> prescriptive; instead, each server should do this for itself as it
> deems appropriate.  Also, I would just inspect the host header
> directly and *never* manipulate the Request-URI.
> 
> <http://tools.ietf.org/html/rfc2616#section-5.1.2>
> 
> The host header is typically used by a reverse proxy or virtual host
> provider to dispatch to the appropriate site handler/module, which
> (usually) should not care what its hostname is.
> 
> >
> > I think there's some link in the inner server module, which dropped
> > 'host' value or created a new uri and throw the old-correct one.
> > Any comments?
> 
> What makes you think that?
> 

Well, I thought 'host' should be kept, and 'read-request-line' works, so
I guess it was dropped somewhere.

OK, if it's not a bug, I think I should avoid to run build-request
again, but modify the original uri an pass the original request. Since
it'll check the validity, it throw error if 'host' is #f. Is that
accepted?

> >
> >
> > Sorry again for the half-baked work, it looks no cool. But I've ever
> > planed a perfect one... :-(
> >
> 
> Lets see it get finished then :-)
> 
> Regards

Re: [Potluck] a lightweight web framework

[Daniel Hartwig]

On 17 February 2013 13:03, Nala Ginrut <nalaginrut@gmail.com> wrote:
> I put here:
> https://gitorious.org/glow/artanis

The examples you mentioned make this look very interested.  Nice job.

Re: [Potluck] a lightweight web framework

[Nala Ginrut]

Hey guys!
Now it's the time to officially release my potluck dish:
https://github.com/NalaGinrut/artanis

Let me re-introduce it:
* just read the README on the link I gave ;-P

Happy hacking!
Enjoy!