Re: Re: Re: Couple of feature requests
"Duncan" <[email protected]>
| Newsgroups | gmane.comp.gnome.apps.pan.devel |
|---|---|
| Message-ID | <[email protected]> |
Pekka Tiittanen <[email protected]> posted 1151748373.5005.5.camel@smack, excerpted below, on Sat, 01 Jul 2006 13:06:13 +0300: > I only want to be able to subscribe to groups on different servers and > download them from that one server only. Over here in Finland there's > usually only one newsserver that can be used and that's the ISP's > server. Using servers from different countries is usually slow so ISP's > server is usually the best choice. So I don't see any good point for > this new arrangement for making primary/secondary/etc servers. Well, one thing I noted awhile back playing with klibido is that if it's just because they are slow, not because one costs more than another, it ends up more or less the way you want it anyway. That is, the slow server doesn't get a /chance/ to download more than say one out of ten parts, because all the others are coming from the fast server. You use up to the limit of your pipe and can't use any more, so the fast server ends up providing almost everything anyway. The slow server only gets in a bit of bandwidth now and then. The same thing happens with completion. Assuming servers more or less the same speed, the server with the lowest completion ends up skipping more parts because they aren't there to download, so gets way ahead of the other one, so you get as many parts as possible from the one with the least completion and naturally pickup from other servers only when the low-completion one didn't have it. IOW, it naturally works the way one would normally want it to work, maximizing the possible efficiency, with no need to artificially constrain things. It follows that the only time one really has to bother with priorities is where one server costs enough more than the others that you want to use it as little as possible, either where it's significantly faster than others available, so must be artificially constrained to keep nearly everything from coming from it, or where you have good enough completion on another server to keep it slow enough that the expensive server would get more than absolutely needed use in the first place. Actually, now that I'm thinking about it, that should apply to your leafnode server as well. If it has the posts predownloaded, it's going to be so much faster than the others that you'll naturally pull nearly everything from it. Thus, there shouldn't be that much of a reason to fiddle with its priority anyway, since it will naturally be the fastest and therefore provide most requests where it has predownloaded them, and naturally slow down dramatically where it hasn't, which would presumably be when you'd want the posts from the other servers. > And I really do want to control where my groups are downloaded from. Now > after thinking this a while, would it be possible to make a per group > setting "download this group only from server" with a dropdown menu of > all the configured servers where to choose from? That'd solve my problem > alltogether. If that feature is added, it'll be post 1.0, because there's no per-group general state retention (only message status, etc) at the moment at all. Whether it'll make 1.1 or any particular version I can't say. In any case, were the feature to be added, it would be a checkbox list, so more than one group could be checked at a time, rather than a radiobutton list or dropdown toggle where only one could be selected at once. > Without that I don't see myself even trying out these weekly releases of > Pan, 'cause they still don't give me anything new or better to test out. That could well be your best choice at the moment. Mainly those that want to download from many servers at once, or are running into the memory scaling issues with the old version, will ATM find the new pan worth switching to. If neither applies to you, the older version is likely your better choice. When per-group state returns, likely with the pre-1.1 betas, things may change, but ATM sticking with the old version is going to be the better choice for many. Of course, eventually the old version will get crufty from lack of maintenance and there will be problems keeping it working as the rest of the world moves forward. However, with luck and Charles' hard work, new pan will be usable enough for you by then (post 1.1 almost certainly, maybe well beyond that, look at the folks, myself included, still running GTK+ 1.x, for xmms, crufty as /it/ is) that changing won't be a big deal. -- Duncan - List replies preferred. No HTML msgs. "Every nonfree program has a lord, a master -- and if you use the program, he is your master." Richard Stallman