W3C home > Mailing lists > Public > www-dom@w3.org > July to September 2001

RE: Memory Overhead

From: Jan-Arve Saeter <jana@birdstep.org>
Date: Wed, 19 Sep 2001 09:40:05 +0200
Message-ID: <9BF5DD8B8F3CD411959500B0D020ED3E2B7098@EXCHANGE>
To: "'Vadim Plessky'" <lucy-ples@mtu-net.ru>, Razvan Costea-Barlutiu <cbrazvan@baltan.bsd.uchicago.edu>, www-dom@w3.org
Somebody have already implied that this is independent of the DOM
specification, 
and to me this seems rather like a problem with the import implementation.

If the mechanism for importing an XML file into the proprieteray internal
structure
is using SAX, (which seems to be a good and widespread practice) then this
can be customized 
by the application programmer in the sense that he writes his own XMLReader
for 
XML text-files. (To make it simple the application programmer should perhaps
use an
existing xml parser)
This is of course more work for the application programmer....

But there's always a balance between usage of human resources and
OS-resources... :-)

;--------------------------------
Jan-Arve Sæther, System Developer
Birdstep Technology ASA
Tel: 24134752

> -----Original Message-----
> From: Vadim Plessky [mailto:lucy-ples@mtu-net.ru]
> Sent: Wednesday, September 12, 2001 11:55 PM
> To: Razvan Costea-Barlutiu; www-dom@w3.org
> Subject: Re: Memory Overhead
> 
> 
> On Wednesday 12 September 2001 16:24, Razvan Costea-Barlutiu wrote:
> |   Vadim--
> |
> |   As a result of the replies i got from the DOM involved 
> people on the
> |   mailing list i figured that the problem is in my implementation.
> |   As Joseph Kesselman pointed out, there are a lot of 
> elements that can be
> |   simply replicated by pointers where duplicates are found.
> |   This is the clear example of a database replication on an 
> XML file, where
> |   the DB structure can be defined in a XSL file or in a DTD 
> and the DOM
> |   implementation can use that information to build the tree.
> |
> |   My DOM implementation (which is at its VERY early stages) 
> simply built
> |   nodes again and again, regardless of the number of 
> occurences of elements
> |   or attribtes, therefore consuming large amount of memory.
> |
> |   I won't post the file because it is a trivial copy+paste 
> created file,
> | for stress-testing only, considering the size of the file. 
> I generated the
> 
> Ok, I see.
> May I ask you to try this file/link:
> http://bugs.kde.org/db/ix/full.html
> (it's exactly what I have on disk, but from Jan.5th, 2001; so 
> current version 
> may vary)
> As I do not meet every day new people doing own (new) DOM 
> implementation, I 
> am very interested in work you done :-)
> So, what will be memory usage/speed of your DOM 
> implementation for URL above?
> 
> | file with XMLSpy from the sample database that you can find 
> in MSAccess.
> | After that, i just Copied and pasted its contents growing 
> exponentially the
> | size of the file. The number of elements reached about 230.000.
> |   So, it's a trivial example.
> |   --Razvan
> |
> |   At 11:11 AM 9/12/01 +0000, Vadim Plessky wrote:
> |   >
> |   >Any chance that you can put this 1.7MB file somewhere 
> online (better -
> |   >zip'ed) so we can test with different browsers?
> |   >Konqueror (www.konqueror.org) has very good DOM-, CSS2- 
> based "KHTML"
> [...]
> 
> BTW: if you are interested in different DOM 
> implementation(s), I highly 
> recommend to take a look on KHTML. It's Object-Oriented, 
> C++-based. Works 
> pretty fast.
> 
> -- 
> 
> Vadim Plessky
> http://kde2.newmail.ru  (English)
> 33 Window Decorations and 6 Widget Styles for KDE
> http://kde2.newmail.ru/kde_themes.html
> Do you have Arial font installed? Just test it!
> http://kde2.newmail.ru/font_test_arial.html
> 
Received on Wednesday, 19 September 2001 03:55:22 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 22 June 2012 06:13:49 GMT