On Tuesday 12 December 2006 15:39, Francis Giannaros wrote:
On Monday 11 December 2006 06:31, Rajko M. wrote:
- Project web page http://www.metalinker.org gives promisses in the same marketing style as menu commercial sites. All is so great, no problems, only benefits, the life is good.
Don't be ridiculous, there's absolutely nothing wrong with shining up your project with a nice website page. In fact that's very good. When I see a new open source project, I'm interested in (i) what it does, (ii) how it helps me.
The problem is shiny doesn't go far enough. I need more details about method, before I can say that it will work for me.
If you want a commercial-like website see: http://basket.kde.org/
...and what's wrong about such sites? Absolutely nothing. You're confusing traits that might be common to open source and commercial apps and then necessarily associating one of them solely with the other
Commercial applications that offer empty shell wave with features and lack details. One (not so good) screenshot and few links to clients is not enough to present one technical idea.
- Downloads are verified for enhanced reliability. -Nothing new for YaST users.
Something new for wget users.
- No Single Point of Failure (SPOF) like FTP or HTTP URLs. -It showed that implementation is not that perfect and single point of failure ruined some people experience. Anything that comes with a lot of declarative sentences and some inserted jokes, can't give me confidence. Jokes alone are not the problem.
That's an issue for you to take on with yourself. Needless to say I wouldn't say this was a common trait.
One of main ideas of metalinks is to remove errors that are not covered with TCP/IP transfer, so if it happens to anyone that download fails because some server has problem, we are back to square one. The problem that you don't want to see is that the number of successes doesn't prove that implementation is good, and one failure proves opposite. It works the same as with software bugs.
- More fault tolerant. -I don't need rsync to repair anything with classic methods like wget.
Rubbish. You can and do.
The wget works fine in normal traffic. It can fail if server (any on the road from source to destination) fails, and that should be covered with metalinks.
As I said, the idea is very attractive, but marketing isn't.
(i) there's nothing wrong with marketing. I'm not sure you really understand the term actually, since SUSE (and every other project practically) has marketing, but just different levels. There is absolutely nothing inherently wrong with marketing, but only the way in which it's delivered. Again, there's absolutely nothing striking about metalink's marketing.
I have no problem understanding marketing Francis, but there is marketing and "marketing". I would like to see metalins idea presented in the way that will satisfy not only people that click on anything clickable, but also those that really read the page before they click.
The download speed and convenient automation of process are not the only factors when someone uses Internet as a software source.
It would be foolish to presuppose that they're not big, or rather BIG, factors. You think I'm going to wait around at release time for a slow torrent or a non-working main site like? Or, you think I'm going to hunt around through a long list of mirrors, when metalink can do it all for me?
The speed and automation are important, but not only things that help me to accept new idea. For me is more important that metalinks clients really work the way they should, and for that I need to know how metalinks is designend. If you have any link to technical reference I would appreciate it.
At release time, now, if I wanted to get an image straight away, there's absolutely no way I'd go for any of the other methods. Why would you? Download is super fast, and if you actually have problems at the end of the download you can use rsync.
The rsync should not be necessary. It is proof that either metalink concept or implementation is not good in all details. BTW, I an use rsync alone and skip any repair.
Needless to say, this point is moot since we've actually since the contrary in our very raw environment. Fedora had very good things to say about it, but there was plenty of first-hand experience for opensuse users in #suse. It was tested, it was proven that it can work very very well.
Regards, Francis.
The Fedora, #suse and any size of positive comments will not help to convince any carefull reader. I can see only black box that sometimes fails on one of it's important features and no one can tell why. I would like that metalinks works perfectly and to be what web site is promising, fine complement to present file transfer options that we have right now. The speed improvement is fine, but now developers have to see why it failed and how to skip rsync phase in the future. -- Regards, Rajko. --------------------------------------------------------------------- To unsubscribe, e-mail: opensuse-project+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-project+help@opensuse.org