Re:And still no Java (key technical leader quit)
on
FreeBSD 5.1 Released
·
· Score: 1
Alexey Zelkin is doing a good job, but needs help since this stuff is way beyond any single person. If the project was more organized and politically clear, there might have been a better effort for corporate out reach, which would have set a a positive example in the BSD for this kind of organization. Currently, you just have old school advocates that don't really understand why these issues are significant (political and technical) and why they need to change, be more politically inclusive and step out of the way of folks when then need to. Linux, on the other hand, is too decentralized to have any politics really effect it. It's definitely more of a meritocracy.
IMO, the open sources BSDs might collapse if this does get corrected. The money being shoved into Linux is enormous.
My time with this project has past. It was kind of a painful experience overall, but I've moved to the Linux community doing RTOS work instead and have found it to be much more reasonable and accepting community. There's really no reason for me to go back to FreeBSD.:\/me wishes the *BSD Java project luck
bill
Re:And still no Java (key technical leader quit)
on
FreeBSD 5.1 Released
·
· Score: 1
It also helps if the FreeBSD Foudation doesn't piss off the main HotSpot VM lead in the FreeBSD. Much of 1.3.1 was mine from the days when I was at BSDi, J2SE 1.4 wouldn't have been possible without work on the threading system and the massive job of porting the HotSpot VM in J2SE 1.3 to FreeBSD. It's a major feat for the BSDs overall and is one of the main bug complains on Sun's site.
Basically, the FreeBSD Foundation paid another lower level engineer to do work that was pretty much in my technical domain without consultation with me asking if this was ok or not, etc.. and other things that go with having the responsibilities of being a technical lead of a project. It undermines the nature of how volunteer organizations work. Pretty ridiculous, since I'm effectively the technical lead of the project until this event. By doing this, they undermined my status within the group, pissed me off and effectively took over the entire FreeBSD Java effort without crediting me for my work over the last 2 years that stemmed originally from my BSDi day as a JVM internals engineer.
I busted ass for these folks the first half of last year and they basically blew me off because of these "elists" actions. It's difficult to interpret it any other way from my point of view.
The flames are thick and I got really pissed off from how they treated me, "you're annoying and threaten my status within the FreeBSD Foundation. Dipshits.
Original archived thread: http://docs.freebsd.org/mail/archive/2003 /freebsd- java/20030209.freebsd-java.html
Supporting emails: http://docs.freebsd.org/cgi/getmsg.cgi?fe tch=42987 9+0+archive/2003/freebsd-java/20030209.freebsd-jav a http://docs.freebsd.org/cgi/getmsg.cgi?fetch=39 387 9+0+archive/2003/freebsd-java/20030209.freebsd-jav a
Reading the CVS log in the patchset will reveal what I've done. An unfortunate set of circumstances.
I stil haven't gotten an apology from them yet, 4 months have past.
Decoding of the instruction set is by far the most intensive of the interpretive process. You'd get at most about 2MIPS with modern x86 CPU driven with that style of engine, so having a strong JIT to get around those issues is absolutely necessary.
I measured this when I was writting DorkCore. The front and back ends of the comiler where pretty easily completed, but the stuff in the middle, intermediate representation, etc... is the real killer for high performance JITs.
You have to do stuff like lazy variable analysis, some kind of SSA (static single assignment) to reduce redundance of writting to a the register file constantly, etc... Putting everything into registers instead of memry increased speed by 50x, but that's still paultry since you have to execute more RISC instructions for the same CISC one, maybe divide by 3x.
I was actually trying to write one myself, but it was simply too hard for any of the traditional UAE 68k techiques, like function pointer tables, to work.
It's a pretty regular instruction set, but the decoding process must deal with a large primary opcode field and then an auxilary field, which make an kind of interpreter execute impossibly slow.
If you remember, I use to be a vMac/Basilisk engineer doing UAE core stuff.;)
What computers still can't do.
on
Arguing A.I.
·
· Score: 1
This is a complicated topic and it's not something that can be shove entirely into a single post.
When you talk about AI, you have to really talk about two different camps. One is symbolic artificial intelligent and the other is neural net AI. Even then you have to talk about the role of philosophical analysis in relation to a science that's trying to provide meaningful structure to organic systems.
1) One being the typical stuff you expect from rule driven systems in Prolog and other AI specific languages designed to deal with exhaustive searching of rules sets. This approach is *dead*.
2) Newer softer approaches (neural nets), although still a tremendeous oversimplification of living systems gets better results but is also dead.
http://socrates.berkeley.edu/~hdreyfus/
Hubert Dreyfus was bagging on this crap in the 60s at MIT and created quite a contraversy against Marvin Minsky, in which Minsky ended up suppressing neural net AI for 15 years because of how they politicized DARPA funding for this kind of research. It wasn't until the PDP group at the University of California at San Diego in the 80s started to publish much stronger results in solving problems like vision, that neutral nets came back and revitalize AI for a break period. Dreyfus, coming from a Heideggerian trained point of view, saw that this is a naive assumption about human knowledge and realized that this was the furthest things from Heidegger's notion of "zuhandlich" (readiness-to-hand), which is his notion of a kind of knowledge that we can implicitly use (like our legs in relation to gravity, you don't think about it) without thinking about it verses "presence at hand" which is the the kinds of knowledges science uses to make sense of the world in explicit terms (lots of thinking about it) within our typical cognitive facilities.
The previous philosophical track where you enframe the world around you through a kind of apriori knowledge, a presence-at-hand, is Heidegger's claim that philosophy has create a lot of artificial problems that don't need to be solved. Problems such as Existentialism, (do we exist ? I think therefor I am ?) is largely bullshit coming from how we seperated ourselves from the world around us by solely believing that our thinking process is what determines our existence. Heidegger restates this as "I am therefore I think" instead and that the previous manner in which we ask the question was based on a false assumption that thinking is devoid of physicality, a religous driven perspective that's originated from Decartes.
Basically, we should focus on the scientific process understanding ourselves through "readiness-to-hand" instead of "presence-at-hand", so that our organic complexity can be described and structured in terms that are meaningful, unlike pure symbolic systems.
We don't think about every mathmatical step when we walk, use gravity, etc...
http://socrates.berkeley.edu/~hdreyfus/html/book s. html
"What computers Still Can't Do: A Critque of Artifical Reason"
His claim that symbolic AI will never work and his reason for this back in the 60s. We now know he's right after repeated failure of that disipine.
Read it. Get educated and drop this geewiz Sci Fi fantasy crap.;-)
Not, true. There are some pretty strong techniques being developed specifically by physical folks that'll make DNS sequencing fairly trivial. It's revolutionary stuff to bio-tech folks, not for typical physic people. bill
I havea buddy that's working on something similar to thing by the name of rabani and UCSD. He was doing some crazy stuff with physics in a lab, but I guess these folks beat him too it. bill
Excellent, I've always wanted PostScript capabilities within X.;-) Floating point coordinates and 2x2 matrices for rotation/scaling too ? Have you looked at QuickDraw GX as technology guide for building the X infrastructure to deal with documents ?
Slashdot Mirror
Alexey Zelkin is doing a good job, but needs help since this stuff is way beyond any single person. If the project was more organized and politically clear, there might have been a better effort for corporate out reach, which would have set a a positive example in the BSD for this kind of organization. Currently, you just have old school advocates that don't really understand why these issues are significant (political and technical) and why they need to change, be more politically inclusive and step out of the way of folks when then need to. Linux, on the other hand, is too decentralized to have any politics really effect it. It's definitely more of a meritocracy.
:\ /me wishes the *BSD Java project luck
IMO, the open sources BSDs might collapse if this does get corrected. The money being shoved into Linux is enormous.
My time with this project has past. It was kind of a painful experience overall, but I've moved to the Linux community doing RTOS work instead and have found it to be much more reasonable and accepting community. There's really no reason for me to go back to FreeBSD.
bill
It also helps if the FreeBSD Foudation doesn't piss off the main HotSpot VM lead in the FreeBSD. Much of 1.3.1 was mine from the days when I was at BSDi, J2SE 1.4 wouldn't have been possible without work on the threading system and the massive job of porting the HotSpot VM in J2SE 1.3 to FreeBSD. It's a major feat for the BSDs overall and is one of the main bug complains on Sun's site.
3 /freebsd- java/20030209.freebsd-java.html
e tch=42987 9+0+archive/2003/freebsd-java/20030209.freebsd-jav a9 387 9+0+archive/2003/freebsd-java/20030209.freebsd-jav a
Basically, the FreeBSD Foundation paid another lower level engineer to do work that was pretty much in my technical domain without consultation with me asking if this was ok or not, etc.. and other things that go with having the responsibilities of being a technical lead of a project. It undermines the nature of how volunteer organizations work. Pretty ridiculous, since I'm effectively the technical lead of the project until this event. By doing this, they undermined my status within the group, pissed me off and effectively took over the entire FreeBSD Java effort without crediting me for my work over the last 2 years that stemmed originally from my BSDi day as a JVM internals engineer.
I busted ass for these folks the first half of last year and they basically blew me off because of these "elists" actions. It's difficult to interpret it any other way from my point of view.
The flames are thick and I got really pissed off from how they treated me, "you're annoying and threaten my status within the FreeBSD Foundation. Dipshits.
Original archived thread:
http://docs.freebsd.org/mail/archive/200
Supporting emails:
http://docs.freebsd.org/cgi/getmsg.cgi?f
http://docs.freebsd.org/cgi/getmsg.cgi?fetch=3
Reading the CVS log in the patchset will reveal what I've done. An unfortunate set of circumstances.
I stil haven't gotten an apology from them yet, 4 months have past.
Decoding of the instruction set is by far the most intensive of the interpretive process. You'd get at most about 2MIPS with modern x86 CPU driven with that style of engine, so having a strong JIT to get around those issues is absolutely necessary.
I measured this when I was writting DorkCore. The front and back ends of the comiler where pretty easily completed, but the stuff in the middle, intermediate representation, etc... is the real killer for high performance JITs.
You have to do stuff like lazy variable analysis, some kind of SSA (static single assignment) to reduce redundance of writting to a the register file constantly, etc...
Putting everything into registers instead of memry increased speed by 50x, but that's still paultry since you have to execute more RISC instructions for the same CISC one, maybe divide by 3x.
It's not a trivial topic at all.
I was actually trying to write one myself, but it was simply too hard for any of the traditional UAE 68k techiques, like function pointer tables, to work.
;)
It's a pretty regular instruction set, but the decoding process must deal with a large primary opcode field and then an auxilary field, which make an kind of interpreter execute impossibly slow.
If you remember, I use to be a vMac/Basilisk engineer doing UAE core stuff.
This is a complicated topic and it's not something that can be shove entirely into a single post.
k s. html
;-)
When you talk about AI, you have to really talk about two different camps. One is symbolic artificial intelligent and the other is neural net AI. Even then you have to talk about the role of philosophical analysis in relation to a science that's trying to provide meaningful structure to organic systems.
1) One being the typical stuff you expect from rule driven systems in Prolog and other AI specific languages designed to deal with exhaustive searching of rules sets. This approach is *dead*.
2) Newer softer approaches (neural nets), although still a tremendeous oversimplification of living systems gets better results but is also dead.
http://socrates.berkeley.edu/~hdreyfus/
Hubert Dreyfus was bagging on this crap in the 60s at MIT and created quite a contraversy against Marvin Minsky, in which Minsky ended up suppressing neural net AI for 15 years because of how they politicized DARPA funding for this kind of research. It wasn't until the PDP group at the University of California at San Diego in the 80s started to publish much stronger results in solving problems like vision, that neutral nets came back and revitalize AI for a break period. Dreyfus, coming from a Heideggerian trained point of view, saw that this is a naive assumption about human knowledge and realized that this was the furthest things from Heidegger's notion of "zuhandlich" (readiness-to-hand), which is his notion of a kind of knowledge that we can implicitly use (like our legs in relation to gravity, you don't think about it) without thinking about it verses "presence at hand" which is the the kinds of knowledges science uses to make sense of the world in explicit terms (lots of thinking about it) within our typical cognitive facilities.
The previous philosophical track where you enframe the world around you through a kind of apriori knowledge, a presence-at-hand, is Heidegger's claim that philosophy has create a lot of artificial problems that don't need to be solved. Problems such as Existentialism, (do we exist ? I think therefor I am ?) is largely bullshit coming from how we seperated ourselves from the world around us by solely believing that our thinking process is what determines our existence. Heidegger restates this as "I am therefore I think" instead and that the previous manner in which we ask the question was based on a false assumption that thinking is devoid of physicality, a religous driven perspective that's originated from Decartes.
Basically, we should focus on the scientific process understanding ourselves through "readiness-to-hand" instead of "presence-at-hand", so that our organic complexity can be described and structured in terms that are meaningful, unlike pure symbolic systems.
We don't think about every mathmatical step when we walk, use gravity, etc...
http://socrates.berkeley.edu/~hdreyfus/html/boo
"What computers Still Can't Do: A Critque of Artifical Reason"
His claim that symbolic AI will never work and his reason for this back in the 60s. We now know he's right after repeated failure of that disipine.
Read it. Get educated and drop this geewiz Sci Fi fantasy crap.
It's being worked on by me and a number of other folks.
;-)
Native threading is in alpha and I'm currently working on the HotSpot compiler.
Not, true. There are some pretty strong techniques being developed specifically by physical folks that'll make DNS sequencing fairly trivial. It's revolutionary stuff to bio-tech folks, not for typical physic people. bill
I havea buddy that's working on something similar to thing by the name of rabani and UCSD. He was doing some crazy stuff with physics in a lab, but I guess these folks beat him too it. bill
Excellent, I've always wanted PostScript capabilities within X. ;-) Floating point coordinates and 2x2 matrices for rotation/scaling too ? Have you looked at QuickDraw GX as technology guide for building the X infrastructure to deal with documents ?