Re: many differents questions

On Mon, 7 Feb 2000, Yanick Gamelin wrote:

> 
> for Jigsaw 2.03
> 
> 1- I want to understand line by line what is in "http-server-index" 
> ,"root.idx" and store file "st-xx", do you have any documentation about-it ?
> can it be understood ?? (this is hard to understand by reading code)

It is exactly the same thing as what you may find in the XML based
serialization.
* http-server-index contains the mapping between the keys and the
  physical location of the stores (relative to the store directory).
* root.idx contains the root container. By default it is also the root
  resource, but it may be changed.
* st-xx is the container xx

Each container contains the serialized resources and their frames, filters
and such. For every child container, there is a key attribute. This
attribute is used to get the store (aka the description of the sons of the
container) of the child container.

In Jigsaw 2.0.* the serialization used was java serialization. In 2.1, it
uses an external serializer which uses XML.

> 2- Do you have a release date for a stable version with XML serialization ???

For now, Jigsaw 2.1.0 with the latest devel release jars should be as
stable as 2.0.x, as soon as the client side cache and servlet 2.2 are
finished, we will publish 2.1.1, then 2.2.0 if noone objects :)
Also, we will continue 2.0.x until people have jdk1.2 available on all
platforms.

> 3- What is consideration, if i want to save information contained in the
> index file and in the store file in a database ??

You just have to change the serializer in the properties:
org.w3c.jigsaw.serializer=org.w3c.tools.resources.serialization.xml.XMLSerializer

See
http://jigsaw.w3.org/Doc/Programmer/api/org/w3c/tools/resources/serialization/package-summary.html
 
> 5- i'm not sure but i thing that unuse store file are not remove from the
> store directory (after deleting a ressource in jadmin, or after reindex)
> that is right and why ???

Some of them are removed, some are not, especially when the server is
killed by ctrl-c or other nasty way of killing the server. The server save
its state (see state in the store dir).
  
> 6- when i add a large directory, (like the workload tree from webbench 3.0,
> see ZDnet.com) it take up to 50 minutes to add it but only 3 to 5 minutes
> with jigsaw 2.01.  After the first indexing, it take only 3-5 to re-index
> the webbench tree. Why it take so longuer to make the first indexing, and
> how to correct this gap.

The frame agregation has changed, at first it was stopping after the first
extension matching a mime type, now it matches all the extensions and adds
all the not already defined attributes (ex you can have
foofile.fr.html.gz.iso-8859-1 indexed as Content_language fr, Content Type
text/html;charset=ISO-8859-1 Content-Encoding gzip. In the previous
version it was only Content Type text/html;charset=ISO-8859-1
Content-Encoding gzip).
Creating resources costs, but after the indexation, as it is in the cache
it won't take that long (just reading the parent container store if it
wasn't already in memory).
However, taking 50mn seems VERY long, there may be improvement possible ;)
Regards,

      /\          - Yves Lafon - World Wide Web Consortium - 
  /\ /  \        Architecture Domain - Jigsaw Activity Leader
 /  \    \/\    
/    \   /  \   http://www.w3.org/People/Lafon - ylafon@w3.org

Received on Tuesday, 8 February 2000 08:43:50 UTC