World: r3wp

will, that is a great article. It explains why the adjacency list 
model is not good enough b/c it's not normalized

wops, I actually meant the" the nested set model" would be a good 
solution, had to write way too much code to retrive, cache, etc with 
the adjacency list model. Still wondering if some good rebol fellow 
has code ready, something like the article but with stored procedures? 
Anything better? mmm, and what if I'd like geolocalization, should 
I buy a book about PostGIS? please tell me there is a easier way! 
thx ;-)

I haven't, but Will, you could be that REBOL fellow ;-)

The new MySQL web site is much less third-parties friendly. It looks 
like MySQL doesn't want to give much attention on alternatives drivers. 
They want ppl to use their own drivers and nothing other. The new 
MySQL protocol licence is more restrictive than before. They described 
a GPLed protocol concept (which a IMO, a very stupid concept...) 
See : http://dev.mysql.com/doc/internals/en/licensing-notice.html

Of course, I didn't use the protocol documentation on the web site 
(due to license restriction and also because, the documentation is 
deeply flawed (intentionnaly??).

Btw, I've released an alpha version of a new MySQL driver implementation 
: complete rewrite from scratch to be fully async under UniServe, 
 optimized decoding in speed (faster code, streaming decoding, etc...), 
supports only servers v5+. You'll find it in the latest UniServe 
archive (see UniServe group for link)

It's event based, so the API is different from the current MySQL 
driver.

Here's a short example :

uniserve/boot/no-loop/with [protocols [MySQL]]

open-port mysql://[root-:-localhost]/test [
	on-login-ok: func [db][
		insert-port db "select * from test"
	]
	on-field: func [db col][
		print ["Field:" col/name]
	]
	on-completed: func [db rows][
		probe rows
	]
]
do-events

The 'on-row and 'on-error events are currently missing.

A query of mine loops and I can not understand the reason.

I gave a look to the protocol and found that :

* defrag-read port pl/buffer std-header-length

returns a 'packet-len of 108544 bytes

the next 

* defrag-read port pl/buffer packet-len


loops because the 'read within 'defrag-read  retrieves only 108543 
bytes.

Any idea on how I can deal with this problem?

Oh... I'm using MySQL 4.1.9 and I have the same problem with both 
0.9.9 and 1.0.7 protocol versions.

I'm aware of only one case where my protocol implementation may fail 
with a size difference of 1 byte between expected size and received 
size: that's the compression header case. This case can only happen 
when the client send a compression flag to the server. My driver 
never send such flag, so this case should never happen.

Options for testing what's wrong :

1) Try to reproduce the problem with the new async driver version 
(see the UniServe package in Uniserve! channel here).

2) Get a packet analyser tool (Ethereal for example) and try to track 
the TCP exchange on port 3306 to see what the server is really sending 
to the client.

3) Upgrade your MySQL server to latest 4.1.x version, or try with 
the latest 5.x version.

4) If all options failed, turn trace/net on, run a test, log all 
exchanges with the server and send it to me by email for analysis.

Just got an idea that should be your first thing to try :

Find the following line in the driver source :

buf-size: cache-size: 10000

and change it to :

buf-size: cache-size: 500'000

Run your tests and tell me if this fix your problem (else try the 
others options).

Great! Changing buf-size to 500000 solved the problem.

Actually what I posted in Chat applies directly to MySQL, if anyone 
happens to know.

So, it looks like a bad read buffer size setting.

The read of the buffer size is done by the 'read-int24 rule, isn't 
it?

True.

Luca, let's try a more cleaner fix for this issue, replace the following 
code :

;--- reading data ---
		if packet-len > pl/buf-size [

   net-log reform ["Expanding buffer, old:" pl/buf-size "new:" packet-len]
			tmp: pl/cache
			pl/buffer: make binary! pl/buf-size: packet-len
			pl/cache: make binary! pl/cache-size: pl/buf-size
			sys-insert tail pl/cache tmp
		]

by :

;--- reading data ---
		if packet-len > pl/buf-size [

   net-log reform ["Expanding buffer, old:" pl/buf-size "new:" packet-len]
			tmp: pl/cache
			pl/buffer: make binary! pl/buf-size: packet-len + length? tmp
			pl/cache: make binary! pl/cache-size: pl/buf-size
			sys-insert tail pl/cache tmp
		]

(just one line changed)

Just a moment...

revert the initial buf-size to 10'000

then try again.

Sure :-)

Going to sleep, will see the result tomorrow. Thanks for your help 
tracking this bug.

Thank to help solve it :-)

Not solved using this last suggestion.

I use the protocol with MySQL 4.1.5 and 5.0.22 and never noticed 
such a problem.

But I'm not storing such a big data

Cal and I modified this line: 

pl/buffer: make binary! pl/buf-size: packet-len

to this:

pl/buffer: make binary! pl/buf-size: packet-len + 10 

and it works now.

Your other suggestion must have fail because tmp may have been zero 
length.

I'm preparing to switch my OS to a slack partition with MySQL 4.1 
(I believe)

OOPS! didn't mean to hit ENTER. sorry. Anyway, there used to be a 
test site at http://powerofq.com- it is no longer available, I believe 
that the test module was 'mysqlprot. are those resources still available?

hi

MySQL Driver 1.1.0 will be released today. The new features are :