[RC5] OGR motivations/achievements in general

Jack Beglinger jackb at guppy.us
Sat Oct 5 10:38:58 EDT 2002


> More features does not necessarily make something better.  Windows has
> lots of features, but most people get tired of rebooting after a while.**

<snip>

Been and done that for now 25 years.  I also now when some one is stock 
programming responses.  

<snip>

> And, Jack you're still paying attention right?, if you really think your
> concept is so good and will further the success of open source projects
> DEMONSTRATE IT.  Make the dnetc guys look like clueless newbs, because you

<snip>

I have all ready... 98% of the code is place now.  And returning packets from 
non-blessed cruncher... you get the picture.  

Two issues of the DNET, follow through... look at the stats change, yep the 
nightly run is faster, but the stats page return is slower and still incomplete.  
The second is security and quality of information.  They know that users 
install DNET and do not enter a email address. Do they fix the client to 
prevent this error?  What about the security of the packets... none there too.  
The argument of a new proxy will make it less secure... Their tongues must 
be deeply planted in their checks.   

Quality of output from DNET is a question... hell, if they keep up their 
existing client maintance... I will not be able to play RC5-72.  The clients for 
some of my machines... are over two years olds!  And that is current official 
one from the website. 

My request for a proxy change is simple and saves DNET resources in the 
long run, plus gives more machine types ability to play, by using my 
resources to handle the some grunt processing and off load (distributed?).  
Also by allowing the proxy to pick random big blocks when it can not get 
data improves the local network load and helps cut the fragmentation and 
duplicate blocks that comes with a large number of crunchers all running out 
of data and all trying to solve the find another block please...

Today the proxy is already trying to reduce the load to DNET by looking for 
duplicate blocks... just take that to next level and have it merge blocks to 
gather that match some constants - same email and OS?  This alone will 
save DNET processing - one packet instead of two or three or...
With that simple of improvement... then the small block local processing is 
basically available.   

We add a touch logic for return same sized packets... The proxy gets 2^40 
from DNET, it trys to return a 2^40.  This single larger packet improves DNET 
processing.  This makes less returns to DNET so stats for a user will wait 
until the large block is returned... but again we are NOT here for the stats :-P

Quality of the stats is question... they return USER, CPU and OS (at least) 
but do not credit stats correctly...  Look at the CPU page of stats...  They 
credit Intel with X86 work... my network is over 1/2 AMD and one is a Cyrix... 
Where are they on CPU stats page.  Also the credit Motorala for IBM PPC in 
AIX machines.  Plus show blocks for machine types ARM and PA Risc, that 
I do not believe have ever run a AIX OS. 

Maybe it is time to listen to others and improve DNET for all.

jackb
--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list