[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: Re: Recognizing the contribution of the developers of XML

• From: Tatu Saloranta <cowtowncoder@xxxxxxxxx>
• To: XML Developers List <xml-dev@xxxxxxxxxxxxx>
• Date: Wed, 30 Aug 2006 14:23:04 -0700 (PDT)

--- Rick Marshall <rjm@z...> wrote:

> One of my favourites ;)
> 
> cpu hogging is a very personal thing. If the
> processor has nothing else 
> to do it doesn't matter - unless you're running on a
> battery about to go 
> flat :(

Yeah, but when parsing XML, CPU always has plenty to
do! ;-D

More specifically, what I am saying is that time saved
in transfer (less bandwidth to use) may be offset by
latency caused by compression codec. So you trade-off
networking overhead to processing overhead. This has
been observed by recent measurements, Wolfgang H for
one can comment more on this.

Of course, if you can compress things ahead of time,
and may not need to decompress immediately, this
tradeoff may work well. So for speculative transfers
(cache/cluster updates) it could work well.

-+ Tatu +-

> 
> Rick
> 
> PS compression works mostly on the tags - see the
> permathread ...
> 
> Tatu Saloranta wrote:
> > --- Paul Topping <pault@d...> wrote:
> >
> >   
> >> That's why there's compression. It works a lot
> >> better than tag
> >> minimization. For one thing, it compresses the
> >> content, not just the
> >> structure. 
> >>     
> >
> > And yet from performance perspective, is much less
> > efficient than tag minimization (or selective
> > compression using somewhat
> markup-specific/optimized
> > schemed); at least if we are talking about general
> > purpose compression like deflate... at least if
> one
> > considers actual throughput -- although it saves 
> > network bandwidth, comp/decompression is a cpu
> hog.
> >
> > Content compression, specifically, often provides
> no
> > throughput improvement at all. Name
> canonicalization
> > and indexing on the other hand has impact as easy
> to
> > implement.
> >
> > -+ Tatu +-
> >
> >
> > __________________________________________________
> > Do You Yahoo!?
> > Tired of spam?  Yahoo! Mail has the best spam
> protection around 
> > http://mail.yahoo.com 
> >
> > !DSPAM:44f50b67214242115153698!
> >
> >   
> 


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

• Follow-Ups:
  • Re: Re: Recognizing the contribution of the developers of XML
    • From: Richard Salz

• References:
  • Re: Re: Recognizing the contribution of the developersof XML
    • From: Rick Marshall