Date: Fri, 7 Aug 1998 21:52:11 +0000 (GMT) From: Terry Lambert <tlambert@primenet.com> To: nate@mt.sri.com (Nate Williams) Cc: tlambert@primenet.com, nate@mt.sri.com, chat@FreeBSD.ORG Subject: Re: Heads up on LFS Message-ID: <199808072152.OAA29583@usr02.primenet.com> In-Reply-To: <199808071604.KAA18835@mt.sri.com> from "Nate Williams" at Aug 7, 98 10:04:10 am
next in thread | previous in thread | raw e-mail | index | archive | help
> > > On the palm pilot, you couldn't allocate 1M of stuff in that short of > > > time. :) > > > > > > (The stuff I'm allocating is about 500-1MB/object, so your point is > > > truly moot. Especially considering the footprint of my entire program > > > is around 60-70MB.) > > > > > > Don't ask like an expert on something you have no idea on. > > > > You shouldn't either > > How much real Java programming have you done? I have reimplemented most of java.* and a lot of javax.* in C++. I am familiar with the interfaces, and, yes, I have done some JAVA programming (I got my JAVA source license back when it was free only a short while after you did). > You have *no* clue what the GC in Java is like. Let's reorganize this tirade a little: > The statement was: > > > JAVA has a nasty tendency to leak like a sieve until the GC hits a > > steady state. As does Modula 3. Let's repeat part of it again, since you appear to have missed it: > > until the GC hits a steady state. Let's repeat it in uppercase: > > UNTIL THE GC HITS A STEADY STATE. Now let's repeat one of your statements: > *worst case* I've even seen in Java is about 2 * 500K-1MB objects not > being GC'd in about 1 second on a heavily loaded (~100% CPU) bound > machine. Sounds like "a steady state"; specifically, it sounds like a steady state of the class "pool retention time based recovery latency". > This is not 'leaking live a sieve', as you so ignorantly put. No, this is 'leaking like a sieve until the GC hits a steady state'; it is *you* who 'so ignorantly' took a portion of my statement out of the context of its limitation criteria to make it look like I was somehow claiming an implementation bug that doesn't exist instead of a language design bug that does (to wit, the reliance on garbage collection). > Do you even own a pilot? I run "xcopilot"; does that count, or do I have to be able to write in grafitti with a stylus for it to count? > Have you even attempted to run a Java VM on it? No. The memory is too small, given the largest possible anount of uncollected memory no longer in use. See my previous posts, and this time, read them instead of merely reacting to them. I've run *compiled* JAVA code (via Greg Hewgill's "JUMP") on the copilot, though. > Do you like making up problems that don't exist, so that you can > look smart? No. I like finding problems that do exist, however. > Do you like changing the subject in the middle of the discussion to > try and be right? No. I like to give specific examples when something I state is a problem is dismissed out of hand by people merely because they have a religious investment in my statement being wrong because it challenges a decision they've already made and their mind is too closed to allow them to reexamine their past decisions objectively. Hopefully, the example will give them the nudge they need to relent and perform the examination. > Never in your statement was there a mention of a palm pilot. The > statement was: > > > JAVA has a nasty tendency to leak like a sieve until the GC hits a > > steady state. As does Modula 3. You're right. This statement shows concern for memory footprint, without identifying the source of the concern. I didn't specifically state "palm pilot" as one of my reasons for concern. > This is catagorically wrong, and shows your ignorance of both Java and > GC techniques in both languages. Obviously I disagree with you (big surprise). > Finally, I have *also* seen Java run on the Pilot. I'll say it again > that it isn't real (yet). Maybe in time, but not yet. The screen/GUI > interface is one of the limiting factors. I'd be interested in where you saw this; I've only seen a bytecode to 68k assembly translator for running JAVA on the palmpilot (or in my case, xcopilot). > You were wrong, and even more so changed the subject in mid-stream and > still were wrong. This is BS. I gave a concern, was bitched at that it wasn't really a concern, replied with a concrete example of why it was a concern for me and *should* be a concern for others. Then I had my throat jumped down by the person who bitched that it wasn't a concern that I was somehow "changing the subject" by virtue of giving concrete examples that contradicted their claims. > You have yet to admit wrongness in any discussion, but its fun to > watch you dance around trying to be right in the process as you > change the subject and the entire focus of the discussion. This is utter BS. I have admitted my errors on both the FreeBSD lists and on Usenet. I admit my errors when I am shown to be in error. Search the archives of both, if you don't believe me. I am actually more forthright in this than most people I know. I have exactly two issues pending that I know about: my probably incorrect characterization of the FreeBSD swap pager, which I have already stated I would have to ask John Dyson about, and that David's statements had made me unsure, and a long Usenet posting that I haven't had time to reply to Caspar Dik about regarding the newer than I was aware of RT capabilities of Solaris (it would be unfair to not correct myself in the same forum in which I made the error; I just haven't had time to write it all out, since I want to be very precise on this one: older Solaris didn't have RT features). PS: Now who is trying to change the subject away from GC'ing being a bad idea? Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199808072152.OAA29583>