| |
For those who are unaware, when Apple released OS 10.2, it broke LWSN, and those who were unfortunate enough to upgrade were awakened to a rude realization that their network renders no longer worked. Whoever's fault it was, Lightwave users were left without the ability to network render.
What OS 10.2 did was enable a file to be read by one program while still being written by another. This is great except that LWSN was reading partially-written command files. This led to commands such as "rende" and "load J". Also, somehow, LWSN would render the command files unreadable after a time.
I have written a program that hacks around this issue (hopefully, at least it has in my testing), and those who are interested may download it at:
http://www.catalystproductions.cc/screamernet/download.html
Please let me know if anyone has any problems with it.
--
-Jon
"This e-mail has been transmitted using 100% recycled electrons."
| |
Amazing. Many people have been wondering why they could not get Screamernet to work at all. Now we know why.
| |
Good investigative skills Jon or should it be Sherlock?, I think you should be on Newteks books, good to see you also look after the environment by recycling! He He He.
Will let you know of any problems.
Thankz.
| |
A real smooth fix! Thank you!!!
| |
How do you use screamer net in the first place? Then, how do you go about applying this fix? I have a brand new "render engine" G4dp 1.25/osx sitting here doing absolutely zilch. Somebody, anybody please excuse my ignorance, but I need help.
Hokeypokey
| |
ok, why is this even an issue?
yeah i know SN is not a big priority at Newtek. But frankly I have been trying to educate Newtek on the Supported interapplication communication methods available to them through Os X for at least a year now, Why isn't someone Cracking the books at newtek?
Is anybody aware that cocoa has a system of interapplication communication? Applescript and Carbon as well? And then theres all of the Unix tools available to os X. There is frankly no excuse for this level of neglect on one of the most important features (take your pick network rendering or batch rendering) of a a flagship product. I read Apple's developer news, its been obvious for quite some time that Old ways (10 + years) of doing things are not going to work, going forward. Apple has delivered a robust opengl platform that gets better at every iteration. Newtek has got to swallow their medicine here, and do some work on LWSN.
| |
[[[Newtek has got to swallow their medicine here, and do some work on LWSN. ]]]
I'm so glad it's not me starting these discussions anymore... I know one thing... Rumor has it that someone is working on a clustering solution for 3D rendering for the Mac a la Rendezvous. As you've stated to me time and again, Ted... I don;t think NewTek will do anything to address the issue. Besides, it works on Windows, so who cares? True networked, clustered rendering could have come to the Mac. Dean Dauger was extremely eager to assist in the development. NewTek had their chance. Now it's likely that the competition will pass them by. The initial poster to this forum stated...
[[[For those who are unaware, when Apple released OS 10.2, it broke LWSN, and those who were unfortunate enough to upgrade were awakened to a rude realization that their network renders no longer worked.]]]
That's pretty funny, isn't it? ;-)
--
Ed
| |
They're just waiting for all of us to get desperate, then they'll unveil the latest version of LW, at the ultra low cost of somewhere near $1500.00!
| |
I don't care if it's a new version of Lightwave or just a patch. As long as it happens soon. There have been so many posts on this subject by people unable to get the Mac Screamernet to work. Now, with Jonathan Baker's efforts we can see why it is broken, and that it's not just user error.
The Mac version of Lightwave is being left in neglect. Total neglect.
| |
worse than that, its not "mac users" that are neglected here... LWSN is based on the available technology of 1989, on any platform. For the longest time, writing to text files was the only way to get a cross-platform network rendering solution to work... but in the last few years we have seen shipped products, with solutions that work, and work going forward. they don't rely on system hacks (referring to the reliance on the system being incapable of reading a file thats being written to here). They are robust and responsible solutions.
Newtek has all the tools available to develop a Robust cross-platform network rendering solution, and Now, I am suggesting that it would take less time to implement than to support LWSN in its current incarnation. I personally don't see anything changing in the next 2 years, but I want my opinion on record here: LWSN is inadequate to its task.
re: clustering:
Clustering is a great technology, but it does not solve newtek's problem. Newtek hasn't kept pace with evolution of the modern OS, and now the "modern Os" (mac and windows) is evolving beyond LWSN. LWSN is dying in a platform agnostic way.
10.2 is a networking powerhouse, with support for All of the major networking protocols: Appletalk IP, appletalk, smb, nfs, etc... It has multiple hooks/frameworks for sending Messages across networks, and a lot of them are cross platform (java, javascript, webObjects, IP) and even if Newtek built their own message system, 10.2 is very helpful in that area. the point is, writing/reading a text file is wrong, no sane programmer would see that as a good thing.
I personally wrote an ftp client in 2 hours in os X (10.2), including research and developement, Using Applescript. And i consider myself a novice/ hobbyist programmer. Its a testament to how understanding the tools can help the design and execution of a project.
| |
A number of issues have surfaced with the release of Jaguar, and the development team is indeed working on identifying and remedying those issues. Aside from the SNet issue, a number of LScript functions seem to be having difficulties, particularly in some of the third party LScripts.
Currently the development team is also reviewing some of the previous forum discussions referred to in this thread, as they consider future Mac development planning.
| |
Well, Chuck, here's NewTek's chance.... I'm sure you could contact Dean Dauger for any distributed / clustered solutions. In this instance, network rendering is somewhat synonymous with clustering. John C. Welch also lurks here from time to time. If it's crossplatform networking or simply Mac networking, then John is the guy you should be talking to. He's also highly proficient with AppleScript. Both Dean and John would probably be more than eager to help NewTek develop a new solution. After all, both of them offered their assistance when the original discussion took place. They were both turned down. Just imagine... A networking/clustering solution for 3D rendering, implementing Rendezvous probably wouldn't be far off at all. If I remember correctly, Dean already has the framework in place. All that's needed is a call from NewTek to start development ;-)
--
Ed M.
| |
Unless they work for free, Ed, then it isn't likely to be as simple as just a call. It's an option to consider, and there would be a lot of things to explore about that option, and a lot of other options to consider before any decisions are made.
| |
Thanks, Chuck, for the response. I think that Ted's comments (above) are also very worthy. OS X now comes with a lot of powerful network features which should be taken advantage of.
| |
As I said, Chuck, I speak with John on somewhat daily basis, he hasn't informed me of anyone from NewTek calling him asking for his assistance. As far as working for free.. Well, I don't know about that. Perhaps you should at least ask what the going rate would be ;-)
Remember, assumptions are usually the mother of all F'ups... Call, if anything is too far out then at least you made an attempt, right? I'm sure the users would agree.
On the other hand, I got the distinct impression that Dean would have helped with the entire project and probably charged very little if anything at all since it would have gained some serious recognition for his company and products. Again, you should contact him and at least show that you are a little interested. Dean would be pleased. You might be surprised at how receptive he is.
I can't speak for John though. Still, you might want to contact them with the specifics. I know both would be happy to assist.
--
Ed
| |
Hi, Ed!
As I said, I've brought the matter back to the team's attention for further review. I don't doubt that calls were not made previously, but there are changes in the team and in oversight of development, and a fresh look at this topic is certainly in order. That review will take some amount of time, and I feel pretty well assured that the available options will be thoroughly explored in this round. Assuming that because we haven't called yet that we are making assumptions that a given option is prohibitively expensive would not be founded.
Chuck
| |
Well that hack finally got things going. At first.
Now however, ScreamerNet will stop at random points during the rendering with the message "Waiting for CPU 1 to begin rendering" even though CPU 1 is listed under the Status column as Ready.
Checking CPU 1, I see the LWSN module just endlessly repeating "Lightwave command: wait". All the nodes in fact are sitting there with the same message.
Attempting to restart ScreamerNet is met with repeated dialogs saying "ScreamerNet rendering in progress. Hit Escape to abort" and of course ESC does nothing.
So no crashes, no node problems, just ScreamerNet ignoring the obvious.
Once again ScreamerNet fails.
- Roland Combes
PerformTech, Inc
| |
That's why you need to use my ScreamerNet Controller :-) That is the sort of frustrations I dealt with to write it in the first place. It's available where the fixer is - www.catalystproductions.cc/screamernet/download.html
Seriously, though, please do check it out, it has been very helpful to all who have used it.
-Jon Baker
| |
Once again, there's an easy way to rewrite SN in a lightweght, cross platform, standards based way...hell, there's two or three:
XML-RPC
SOAP
other Web Services
These are not idle theories, they are doing work *today* and are well understood. Proper implementation of this would make SN a far more powerful application, yet it would still be simple enough to avoid problems.
Face it, once you get away from the wretched setup, SN ain't exactly rocket science, so moving it to a different model would be pretty simple.
john
| |
It's sad these OSX technologies are not being taken advantage of. I think within Newtek they must realize that the current implementation of Screamernet is useless and needs replacement with something better.
| |
This was taken from the Architosh website:
Reader Feedback: Apple's Users Demand More, where's the Pooch?
An Architosh reader wrote in regarding our article on the Pro Mac Survey: Apple's Users Demand More.
Hi there,
There was no mention of: http://daugerresearch.com/pooch/whatis.html
which has been tested with an over 50 Mac cluster with great results. This is the kind of things that people want for computing power...with regards,
Lachlan Deck
Yes, the ability to do parallel clustering is currently available to advanced Mac users willing to adventure with a group of Macs and some numerically intensive apps. Like the UCLA Mac cluster supercomputing project, this link to Pooch will get you up and going with your own parallel supercomputer. For Pooch screenshots visit this page.
However, it is unclear to us what Mac 3D apps could be made to work with Pooch and parallel computing. Any suggestions from you readers? We have contacted Dauger Research Inc. and have put in a request for information. Architosh would be willing to help sponsor any effort to create a "Pooch-aware Mac 3D application" (specifically targeted at rendering) and would give a software developer free advertising space if they donated resources for the project.
If you are interested and/or have more information please let us know.
Here is the URL:
http://www.architosh.com/news/2002-03/2002b-0327-pooch.phtml
--
Ed M.
| |
Beam,
Apologies, but the fact that you and others would like something better does not make the statement that the current version of ScreamerNet is "useless" true in even the least degree. Thousands upon thousands of users are rendering with it on all LW platforms, both by itself and supplemented with third party render controllers, and as such it is greatly useful indeed.
Is something easier to use and more flexible possible? Yes.
Are we reviewing the possibilities? Yes.
| |
Hi Chuck. Well, my statement was based on the fact that my main machine is a new Dual GHz G4 Mac which came with OSX 10.2 installed. I can't go back to earlier versions of the OS. Screamernet, as it comes from Newtek, will not work at all with these new Macs.
I'm sure there are thousands of happy Screamernet users rendering away to their hearts' delight. However they're not owners of new Macs!
I am very happy to hear that Newtek is reviewing its implentation of network rendering. I look forward to the arrival of these new technologies, and hope it takes full advantage of OSX's advanced newtwork features.
| |
Hi, Beam!
My first message on this thread mentioned that we are working on addressing the issues that have arisen with the release of Jaguar, which includes the issues with SNet; please rest assured that a fix to that issue is an item unto itself, and would not await the more comprehensive review that a decision to re-implement SNet to other networking standards will require.
Chuck
| |
Chuck:
"please rest assured that a fix to that issue is an item unto itself, and would not await the more comprehensive review that a decision to re-implement SNet to other networking standards will require. "
First of all, as was pointed out to me, SNet has nothing to do with networking other than you have to create high-overhead user - level mountpoints so standard file calls that only function on local filesystems will work.
Second:
"Apologies, but the fact that you and others would like something better does not make the statement that the current version of ScreamerNet is "useless" true in even the least degree. Thousands upon thousands of users are rendering with it on all LW platforms, both by itself and supplemented with third party render controllers, and as such it is greatly useful indeed."
Is Snet useless? No
Is SNet so painful to set up that given a choice between a boil and setting up SNet, the boil doesn't lose by much? Yes
By that token, I could say that useing assembly code to render is useful, since it's possible. and that's about where SNet is...it's possible to use it, and if it happens to be St. Swiven's day, it may even be not hard to use. But from what I've seen, you'd be *hard* pressed to find anyone who views SNet as much more than a boil on their bum.
The fact that SNet is not considered anything other than painful to use without third party addons should have been your incentive long ago, especially since the methods I've pointed out work across every platform that LW runs on.
john
| |
Thanks, Chuck, I'm glad to know that the issues are being considered.
| |
I'd like to make a request here.
Please -- I'm trying to address these very issues, and need a little help.
Please, provide me with a step-by-step on a reproducable problem - hopefully with a single mac.
I'm new to the mac, but really want to help out here.
Please, email me at deuce@newtek.com with any assistance you can provide.
~~Deuce
| |
I just contacted deuce -- I am running 10.2.3 and am unable to reproduce the problems described. As far as useability, and ScreamerNet, I'm not convinced there are significant problems with the application. Sure, there could be a better controller built into LW. Sure, SN could have a prettier interface. So what? If you spend all your time rendering, I guess it's a big deal. I spend a lot more time modeling, animating, and surfacing. When it comes time to render, I allot the 20-odd minutes to get everything organized, and set it up.
I've run successful render farms on networks from 2 - 15 macs over the years, with LW 5.6 - 7.5 Maybe it's my longer term working with SN, but it doesn't seems so hard. On top of that, there are literally HUNDREDS of tutorials online on setting everything up, so there's no real excuse about the simplified documentation. I'd rather get freee unlimited nodes that require a little work than one or two nodes with a pretty interface...