[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Why XPath?



       Val Iliescu wrote:
> 
>         I am really trying only to understand your architecture:
> 
>         Why is everything ( I mean Pub/sub matching ) concentrated into 


> the DOM/XPath ??
>         I see the matching as a fundamental problem ( algo ), and it 
> not be deferred to     XPath ???
>         I mean how can u be performant with such an open 'key' kinda 
> design ?
>         The guys at INRIA have *really* concentrated into how to match 
> clients to       subscribers lists and it seems to be warranted ...
> 
>         Why do I bother ? because I liked the rest of your stuff (aka 
> callbacks...),
>         but it seems to me you stopped short of the rest ...
Hi Val,


I believe using XPath as our query language,
to query over xml based meta informations (topics) is one
of the prime design desicion in xmlblaster.

You can choose using the exact oid to access a message or use
XPath to retrieve the messages you desire.

In most use cases, XPath query should not be a problem.
Imagine an air traffic controller which queries all wind speed
devices on his runways. His software subscribes once
to "//WindSpeed" early in the morning, and thereafter the 
Windspeed is automatically updated in real time the whole
day long.

So XPath is no performance bottleneck at all.
A pilot doing his pre flight briefing in Tokio,
may want to query the Windspeed in Peru (his destination),
so his software could query something like this:

    //Airport[ at icao=\"LIMA\"]//WindSpeed

No problems with XPath.

Manuel did some performance test, and for huge data,
XPath performance did beat Regex performance
so this enforced our design decision.

regards 

Marcel


PS: Where do i find informations about the INRIA approach?