Seti clients should be open source

[jmeadlock]

Here's a page describing some history:

Free=good

<RANT>

I'm new to SETI@Home but I'm quite irratated because I believe the client software is very poorly optimized (though I have no proof because I'm not an assembly guru and don't have the source code to look at).

Why would fixing a buffer overrun exploit make the client 5-10% slower? Does the client take advantage of MMX/SSE/SSE-2 extensions? (I don't think so... but I don't know if it could make use of them either). Why isn't there a version for Opteron's or Itanium's? Laziness? If the source was public all of this could be answered and the clients would be much more efficient.

The FAQ states they decided to make the code proprietary "for security reasons and for science reasons as well".

I can't concieve of a scientific reason to hide the code; I assume they think it would be more difficult for them to ferret out cheaters if the code was free. I'll bet they'd have an army of coders in line to help them with security concerns if the code was available.

My main gripes are that their software isn't available for the community to look at and improve, and I don't believe they're making efficient clients; they could go to a scheme like:

1) publish source code
2) release binaries for 1000's of configurations (instead of their current 60 or so) with a key built into them that isn't included with the source to help eliminate cheaters (cheaters can always modify the binary)
3) accept contributions from whoever that improves the code.

Mail these guys if you agree:

davea@ssl.berkeley.edu
danw@ssl.Berkeley.edu

</RANT>

Edits: My crappy spelling/grammar

 

[Crazee]

Remember that Seti was originally setup 3+ years ago. I believe that the BOINC stuff they are working on will be open source. There really is no need at this point to open the source code up on a project that has a definite ending timeframe.

 

[jmeadlock]

Remember that Seti was originally setup 3+ years ago.

Ok.. and it was free software then.

I believe that the BOINC stuff they are working on will be open source.

I hope so.

There really is no need at this point to open the source code up on a project that has a definite ending timeframe.

Why not? It could improve efficiency and perhaps improve their next client.

(A reasonable argument I can think of is their time would be much better spent on the next client than trying to deal with the new security problems they'd introduce by publishing the source... playing devil's advocate to some degree here).

 

[Confused]

I agree that it sucks that the client hasn't been MMX/SSE/SSE2/3D-Now!/Extended 3D-Now! enhanced, as it would help improve the efficiency of the client no end!

However, releasing the source code for such a scientific experiment is not a good idea. There are many people out there who could easily figure out the code, and work out how best to be able to manipulate the results. This could be as simple as getting it to "finish" a WU within a couple of minutes, or a few seconds, to submitting totally useless results, and rendering the entire project worthless. They could also find the code that enables it to send/recieve to/from the server, and possibly figure out how to manipulate the server. I'm not a hacker or programmer, so I don't know exactly what/how, but it is possible.

Originally posted by: Berkeley
BOINC is open source (though applications need not be). If you want to customize your client, or fix a bug in it, now you can. The server-side parts of BOINC use other open-source software like Apache, PHP, and MySQL.

Looks like BOINC will be open source, though it seems the actual applications won't be (which is good, IMO). The actual code for the projects won't be open source, but the "handler" will be, of course we shall have to see how well this does or doesn't work.

I also agree that there shouldn't be a 5-10% slowdown down to purely security fixes, my guess is that they have slowed it a bit, so that their servers don't have as much traffic/bandwidth used, although the 100mbit pipe they got last year should have more than helped that.


Confused

 

[Assimilator1]

Why isn't there a version for Opteron's or Itanium's? Laziness?

Jesus give them a break!

,they are only a handful of volunteers who run SETI@H ,seeing as the Opteron has only very recently been released & both the Opteron & Itanium represent a very small number of the cpus that would be running S@H I hardly think that is a high priority.Not to mention they've been busy with BOINC & the cheaters in S@H1

I do agree though that v3.08 shouldn't be slower though!:frown:

 

[networkman]

Personally, I like the "closed-source" client as it is; and yeah, I'm one of the many folks out there who isn't a big fan of open-source code. I like canned code - I like getting my programs from a very limited number of vendors.

Something about having too many cooks in the kitchen comes to mind..

EDIT: Another thing - yes, I do know the value of shareware and such. But as for "FREE" - the old addage: "You get what you pay for." rings true for this guy.