VNC server based on kdrive using damage extension?

[Kurt Pfeifle]

> VNC server based on kdrive using damage extension?
> Michael L Torrie torriem@chem.byu.edu
> Wed, 18 Feb 2004 09:40:16 -0700
> 
> 
> On Tue, 2004-02-17 at 18:23, Kurt Pfeifle wrote:
> 
>> I am using mozilla, KDE Desktop, OpenOffice, KMail and other
>> programs over a 56k modem connection on a daily basis  --  it
>> doesn't even consume all the bandwidth! I guess  20 - 40k
>> would suffice for most of the time.
>> 
>> However, I don't use plain X. I fell into the beautiful NX
>> trap offered by http://www.nomachine.com/ and can't live
>> without any more.
> 
> How does NX compare to simply using SSH to compress the tunneled X
> traffic? 

NX is better on the compression front alone (albeit this is not
by orders of magnitude, of course).

But it really shines because it implements 2 other techniques
to make it so efficient:

  * eliminate roundtrips almost completely
  * uses a cache to save even more data transfer than a pure
    compression would.

> If NX can safely eliminate round trips to the server,

It does....

> then I'm
> all for it.

...so welcome to the club!

> Why is it not a standard part of X11 if it can do such a
> good job of eliminating round trips and compression? 

I really don't know. Too little people did care to testdrive
it. Too many people (even well-known hackers) spread rumours
or hearsay without any own experience with it. (I always
experience Deja Vu-like flashbacks when I read their
ill-informed comments: I've been through it with CUPS vs.
LPD/LPRng years ago. Sometimes I can't decide if this is
tragic or very funny....)

But, to get back to the real question: why make it a "part"
of X11 at all? It works perfectly well the way it is
architectured now -- as an additional layer, which leaves all
legacy libraries, applications and X servers alone, but let
them work greatly through the NX tunnel.

To make NX *part* of X11 would probably require re-compiling
of all apps and libraries and X servers. And not everybody
could enjoy the fruits of NX development as they can now.

It is really only a matter of 3 minutes *now* to hava an NX
client or an NX server up and running and work for all apps
and desktop types....

The way it is now: you can use it without having it made as
part of X11. Guys who don't like it, leave it alone. But still,
its performance and many fields of deployment (building a
really viable competition to Citrix and Tarantella and SunRays,
making Linux desktop migrations a lot smoother for large
enterprises or government bodies, etc.) will ensure that
it will come to rescue a long-term perspective for X-based
software development (shielding off attempts to create
alternative foundations of GUI -- or leaving them much more
time to become mature enough, if you like....)

> Any remote X
> displays (even on a fast network) could benefit from that.
> 

Yes. Very true.

I have now some of my printing customers (who saw NX
by accident on my laptop, and how I connected from Germany
to a desktop in Italy) testing it in their LAN. They can
now save the cost of an expensive hardware upgrade (having
a 10 MBit net, with many parts not switch-ed but HUB-ed)
to a 100 MBit switched ethernet.

> Michael
> 

Cheers,
Kurt

> 
>> 
>> Mozilla 1.6 over a 9.6k GSM Modem link requires 5 minutes to
>> startup and takes 4.000 roundtrips if I use plain-ol' X.
>> Unusable -- true.
>> 
>> Same Mozilla 1.6, again over a 9.6k GSM Modem link, but this
>> time turbo-powered by NX differential compression takes only
>> 20 seconds to startup and sends only a dozen roundtrips if
>> across the wire, ahem ether.
>> 
>> KDE-3.2 to "boot" transfers 4.8 MByte of data until the
>> desktop is there (after KDM login), if running through
>> vanilla X.
>> 
>> Same KDE-3.2 to boot across the "NX wondertool" transfers
>> only 38 kBytes of data.
>> 
>> All numbers taken 30 minutes ago.
>> 
>> HTH,
>> Kurt
>> 
>> > -- Matt Jones
>> > 
>>