You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Ian Holsman <ia...@cnet.com> on 2001/11/26 23:49:20 UTC

2 modules

I've got 2 modules which I believe would be of use to others, and could
go into experimental.

1. mod_ssi_stub is of interest to module developers. It allows people to
stub out custom SSI tags with a bit HTML code. This has helped us when
other developers haven't had their modules ready in time.

2. mod_sentinel (bad name) skips the file/directory walk and does a
open/fstat instead of stat/open to read the file, resulting in a
performance win when the file is on NFS, and you don't need to check
file permissions etc

are these 2 worthy of 'experimental'?

-- 
Ian Holsman          IanH@cnet.com
Performance Measurement & Analysis
CNET Networks   -   (415) 344-2608


Re: 2 modules

Posted by Justin Erenkrantz <je...@ebuilt.com>.
On Tue, Nov 27, 2001 at 08:42:45AM -0800, Ian Holsman wrote:
> On Mon, 2001-11-26 at 21:09, Justin Erenkrantz wrote:
> > On Mon, Nov 26, 2001 at 02:49:20PM -0800, Ian Holsman wrote:
> > > 1. mod_ssi_stub is of interest to module developers. It allows people to
> > > stub out custom SSI tags with a bit HTML code. This has helped us when
> > > other developers haven't had their modules ready in time.
> > 
> > Can you give an example of its usage?
> > 
> SSI Stub SSIcmd "<blink>TBD</blink>"
> 
> this way when you have a team of developers they don't need to wait on
> each other, and you can benchmark/debug/perf test using the same HTML.
> It isn't meant for production use, just for development

I'm still missing something.  What does the file look like before this 
module and what does it look like after?  -- justin


Re: 2 modules

Posted by Ian Holsman <ia...@cnet.com>.
On Mon, 2001-11-26 at 21:09, Justin Erenkrantz wrote:
> On Mon, Nov 26, 2001 at 02:49:20PM -0800, Ian Holsman wrote:
> > 1. mod_ssi_stub is of interest to module developers. It allows people to
> > stub out custom SSI tags with a bit HTML code. This has helped us when
> > other developers haven't had their modules ready in time.
> 
> Can you give an example of its usage?
> 
SSI Stub SSIcmd "<blink>TBD</blink>"

this way when you have a team of developers they don't need to wait on
each other, and you can benchmark/debug/perf test using the same HTML.
It isn't meant for production use, just for development


-- 
Ian Holsman          IanH@cnet.com
Performance Measurement & Analysis
CNET Networks   -   (415) 344-2608


Re: 2 modules

Posted by Ian Holsman <ia...@cnet.com>.
On Mon, 2001-11-26 at 21:17, William A. Rowe, Jr. wrote:
> From: "Justin Erenkrantz" <je...@ebuilt.com>
> Sent: Monday, November 26, 2001 11:09 PM
> 
> 
> > On Mon, Nov 26, 2001 at 02:49:20PM -0800, Ian Holsman wrote:
> > 
> > > 2. mod_sentinel (bad name) skips the file/directory walk and does a
> > > open/fstat instead of stat/open to read the file, resulting in a
> > > performance win when the file is on NFS, and you don't need to check
> > > file permissions etc
> > 
> > Wouldn't it be better to have this as an option in the core 
> > than as a separate module?  -- justin
> 
> ++1  on integrating an apr_file_t * right into the request_rec.  This would
> be an enormous advantage to everyone.
> 
> But don't suggest it's a user option - it's not.  And in fairness to our
> module authors, lets please table it till 2.1.  Simply, we try a direct hit
> on the file.  If we succeed with open/fstat, great, if not, we do the usual
> stat [must be a dir, or access was denied.]
> 
> Other modules will win, too.  A file permanantly opened [cached] becomes 
> nothing but a dup2.  mod_mime_magic doesn't have to add an extra open/close
> to the mix, it's nothing but a read/seek(0).  This is generally goodness.
> 
> But I'm guessing it's 3 months before everything is optimized and settles
> out from this [very worthwhile] change.  Do we make everyone wait three more
> months?  No, IMHO we offer a savory performance enhancement with 2.1.
> 
agreed I'd rather get 2.0 than wait 3 months.. so this will be a 2.1
mod.
> Bill
-- 
Ian Holsman          IanH@cnet.com
Performance Measurement & Analysis
CNET Networks   -   (415) 344-2608


Re: 2 modules

Posted by "William A. Rowe, Jr." <wr...@covalent.net>.
From: "Justin Erenkrantz" <je...@ebuilt.com>
Sent: Monday, November 26, 2001 11:09 PM


> On Mon, Nov 26, 2001 at 02:49:20PM -0800, Ian Holsman wrote:
> 
> > 2. mod_sentinel (bad name) skips the file/directory walk and does a
> > open/fstat instead of stat/open to read the file, resulting in a
> > performance win when the file is on NFS, and you don't need to check
> > file permissions etc
> 
> Wouldn't it be better to have this as an option in the core 
> than as a separate module?  -- justin

++1  on integrating an apr_file_t * right into the request_rec.  This would
be an enormous advantage to everyone.

But don't suggest it's a user option - it's not.  And in fairness to our
module authors, lets please table it till 2.1.  Simply, we try a direct hit
on the file.  If we succeed with open/fstat, great, if not, we do the usual
stat [must be a dir, or access was denied.]

Other modules will win, too.  A file permanantly opened [cached] becomes 
nothing but a dup2.  mod_mime_magic doesn't have to add an extra open/close
to the mix, it's nothing but a read/seek(0).  This is generally goodness.

But I'm guessing it's 3 months before everything is optimized and settles
out from this [very worthwhile] change.  Do we make everyone wait three more
months?  No, IMHO we offer a savory performance enhancement with 2.1.

Bill


Re: 2 modules

Posted by Justin Erenkrantz <je...@ebuilt.com>.
On Mon, Nov 26, 2001 at 02:49:20PM -0800, Ian Holsman wrote:
> 1. mod_ssi_stub is of interest to module developers. It allows people to
> stub out custom SSI tags with a bit HTML code. This has helped us when
> other developers haven't had their modules ready in time.

Can you give an example of its usage?

> 2. mod_sentinel (bad name) skips the file/directory walk and does a
> open/fstat instead of stat/open to read the file, resulting in a
> performance win when the file is on NFS, and you don't need to check
> file permissions etc

Wouldn't it be better to have this as an option in the core 
than as a separate module?  -- justin