FYI: Creating circular references is a perfectly OK thing to do.
in [JavaScript]
|
From: Scott Sauyet on 29 Jul 2010 14:10 Ry Nohryb wrote: > On Jul 29, 2:02 pm, Scott Sauyet <scott.sau...(a)gmail.com> wrote: > >> (...) Others in this thread have rightly argued that circular >> references with host objects is a bad idea regardless of IE's >> particular issues. (...) > > Is that a saying ? Why is it a bad idea? Because they (well, it's been > only Asen) have said so but have given no proof nor reason for it to > be so, except that IE leaks. Because it isn't bad per se. CRs aren't > neither good nor bad. It's just that the word has spread that CRs leak > memory, which is false. IEs are what are a bad idea, not the CRs. Perhaps it wasn't well argued earlier, and I don't have time to check, but it's a simple corollary of the fact that there are few standards that apply to host objects. They are allowed to be implemented however they want. There is no standard I know of that MS is ignoring by its separate GC for host objects, so even if you manage to single- handedly chase IE out of the market, the next great browser might have similar problems. -- Scott
From: Ry Nohryb on 29 Jul 2010 16:42 On Jul 29, 8:10 pm, Scott Sauyet <scott.sau...(a)gmail.com> wrote: > > Perhaps it wasn't well argued earlier, and I don't have time to check, > but it's a simple corollary of the fact that there are few standards > that apply to host objects. ISTM that a GC should release asap the memory used by any object that is no longer reachable, and all browsers except IE manage to do it right, and I don't see any connection between this and the number of standards that apply to host objects. The only connection I see is with the quality of engineering of the software: defective and non working: Microsoft's IEs, ok/working: everybody else's. > They are allowed to be implemented > however they want. There is no standard I know of that MS is ignoring > by its separate GC for host objects, so even if you manage to single- > handedly chase IE out of the market, the next great browser might have > similar problems. You point being that somebody else might as well deliver another browser as broken and badly engineered as Microsoft's Internet Explorer? I'd find it hard to believe... :-) I think that M$ could do much better if they wanted, but they've just not wanted for more than 10 years, and we've been suffering it for long enough. Or not? -- Jorge.
From: Ry Nohryb on 29 Jul 2010 17:20
On Jul 29, 7:10 pm, Stefan Weiss <krewech...(a)gmail.com> wrote: > On 29/07/10 16:17, Ry Nohryb wrote: > > > It's hard to tell what worries [Steve Ballmer] the most when he's not > > on cocaine. [...] > > The ~ null penetration in web servers ? > > That's not even close to reality:http://news.netcraft.com/wp-content/uploads/2010/07/wpid-overallc.png ~ 20% is a ridiculous server marketshare for Microsoft. They try hard to increase it, but they fail again and again. Nobody sane prefers an IIS over a Linux server. > > Linux? > > Are you sure?http://gs.statcounter.com/#os-na-monthly-201001-201007 I was thinking about Linux's share in the server market. > > The plummeting market share of IE ? > > Are you sure?http://gs.statcounter.com/#browser-na-monthly-201001-201007 Yes I am. IEs had more ~ 90% market share only 5 years ago... what is it now? > It's true that MS are losing their monopoly in many of the areas you > stated, but they're not doing nearly as badly as you think. Wishful > thinking, perhaps? No. M$ is the biggest zeppelin in the sky, but it's slowly leaking hidrogen, they've got to manage to fix that or sooner or later... Recently Apple's market value surpassed Microsoft's. Can you tell me a single Microsoft bussiness that's not going downwards ? Office ? Window mobile? IIS ? Zune? Bing? Tablets? what? Yes, they've said that sales of W7 are going well. We'll have to trust it. But still, overall... don't you see that everything else isn't ? > >> (...) Others in this thread have rightly argued that circular > >> references with host objects is a bad idea regardless of IE's > >> particular issues. (...) > > > Is that a saying ? Why is it a bad idea? Because they (well, it's been > > only Asen) have said so but have given no proof nor reason for it to > > be so, except that IE leaks. Because it isn't bad per se. CRs aren't > > neither good nor bad. It's just that the word has spread that CRs leak > > memory, which is false. IEs are what are a bad idea, not the CRs. > > Circular references aren't bad per se. In certain data structures, > they're even required. It's true that they present a problem for > reference counting garbage collectors, but there are strategies for > working around that (sweeping/tracing). IE, with its dependance on > external COM objects, has a lot more trouble breaking these reference > cycles. There have been some improvements in IE7 (and patched versions > of IE6, IIRC), but the leaks are still a lot more pronounced in IE than > in other browsers. I wonder if they've finally managed to tackle this > problem in IE9. It's going to be very embarrassing if they don't. > In the meantime, we can either avoid creating circular references where > possible, or break them ourselves - e.g. by nulling variables or > properties, or by calling faux destructor methods when a problematic > object is no longer required. Yes, all of you've been doing it for years. Aren't you tired of it? > >> But the main point is that there are techniques > >> almost as easy to use that don't cause problems in what is still the > >> most widely-used browser. Why not use them? > > > Because there's no need. Because there are much more interesting > > things to code and think about than workarounds for Microsoft IEs' > > bugs. > > If you can afford to ignore IE's memory issues, you're leading a charmed > life. Many of us have to deal with this and other IE-related problems in > our jobs every day. I can't just tell my clients to switch the whole > company over to a different browser, just so that some little script of > mine doesn't cause memory leaks. They'd probably find it more economical > to switch to a different JS developer instead. Yes, I understand that any customer is good, even those browsing with IE :-( > > And because we should not circumscribe the web to the least common > > denominator (for cross browserness) when the least common denominator > > is a fraction of what it would be if it were not for IE. > > I sympathize with your position, I really do, but the fact of the matter > is that most people in this business can't afford to ignore IE. All we > can do is create awareness of the problems, and create enough bad > publicity for MS to force them to do something about it. It looks like > they're finally starting to get their act together with IE9, but it's > still too soon to tell. And we'll still have to live with the older IE > versions for quite some time. I hope not for too long. > That said, there are situations where it's possible to wean people and > even smaller companies away from IE6. One strategy that I've employed > for this purpose is to calculate the price for an intranet project as > usual, and then offer a significant discount if they agree to ditch IE6 > in favor of a more standards compliant browser. The discount doesn't > cost me anything, because I'm saving a lot of time in development, and > now the company has a financial incentive to finally upgrade their browsers. Very good. +1 :-) -- Jorge. |