| Forum Home | ||||
| PC World Chat | ||||
| Thread ID: 81130 | 2007-07-17 12:06:00 | Yet another thread about the forum clock... | Erayd (23) | PC World Chat |
| Post ID | Timestamp | Content | User | ||
| 569777 | 2007-07-18 23:29:00 | Just put Tardis 2000 on the server and be done with it!Tardis appears to be a Windows program. The PressF1 server runs Fedora Linux... | Erayd (23) | ||
| 569778 | 2007-07-19 03:00:00 | Bloody hell . "Correct time stamping of postings" is hardly essential . I have seen time-related problems here . Once . The time was set back, so the "same time" was occuring twice . This affected the sorting of postings for that period . :( (Perhaps the primary key for postings should be a sequence number . ) The important thing about the server time on this forum (or any system using a database which uses time as a key) is that it should be monotonic . "Server updates" will fail? Software updates are done on the basis of version numbers, not dates or times . They can't be organised sensibly any other way . Backups? Where is the problem? MS put the "Archived" status bit into the file system for use in backups . . Reconciling firewall logs with server logs . Trivial . And the firewall probably has the same time . as the web site server . Some types of server clustering . . . all clustered boxes will have the same time . Who would not do that? The absolute cluster time is irrelevant . . . Kerberos authorisation will not be bothered . . . the K server will be in the same LAN, with the same time . As for cookies, don't cookies contain the server's time as a data item inside the packet? The server is passed the cookie by the browser . I doubt very strongly whether the server has access to the user file system timestamp at all . That could be a major security problem . The user system timestamp on the file should be used only for the user system to remove cookies after some time . That's how I would programme cookies . I wouldn't be stupid enough to do that sort of thing in a way that relies on the time setting on thousands (or millions) of other computers (which I have no control over) . The server knows its own time "now" . It can compare that with the timestamp from a cookie it has just been given . Asking for the user's current time would be stupid . The server has better things to do . Also, as I pointed out, there are a hell of a lot of computers showing the wrong time . Many will be days, months, years out . Many (on local networks which get their time set from the local server) are one hour out for weeks after a daylight saving time changeover . So much for your cookies experience, vinref . Look for some other reason . I'm not convinced . The time on this forum can be a long way out causing no problems except for those haven't got watches, or who can't see the sun . |
Graham L (2) | ||
| 569779 | 2007-07-19 04:03:00 | This is great, in fact to keep my entertainment level up, we might just even make the clock even more wrong. this is like Vegas, baby, we don't want you guys to be confined by silly things like time, we want you all to get lost in the wonderful, wonderful world that is pressF1. |
Jan Birkeland (4741) | ||
| 569780 | 2007-07-19 04:16:00 | Bloody hell... Indeed. Does anyone else suspect Graham is just baiting us? |
vinref (6194) | ||
| 569781 | 2007-07-19 04:48:00 | Can you make the clock be the correct time when you login, then freeze. Then maybe some cool javascript when you mouse over it it jumps to the side/up or down. |
Rob99 (151) | ||
| 569782 | 2007-07-19 05:05:00 | The clock should run on time, it just looks amateurish if it doesn't. For heavens sake can't a computer help forum have a clock that runs on time? Computer geek version of "she'll be right". |
zqwerty (97) | ||
| 569783 | 2007-07-19 05:31:00 | Bloody hell.Point noted. "Correct time stamping of postings" is hardly essential.No-one ever said it was; correct timestamping is a convenience thing only. Most of us like our clocks correct. Perhaps the primary key for postings should be a sequence number. The important thing about the server time on this forum (or any system using a database which uses time as a key) is that it should be monotonic.The primary key is a sequence number. The posts are only sorted on a time index (not a key), what the primary key actually is is irrelevant provided it is unique. "Server updates" will fail? Software updates are done on the basis of version numbers, not dates or times. They can't be organised sensibly any other way.Your are correct in stating that the updates are done on version numbers, however any software update that uses tar (and that's most of them) will be affected - as tar will, by default, simply refuse to extract anything with a modification time in the future. Backups? Where is the problem? MS put the "Archived" status bit into the file system for use in backups..Despite your obvious personal preference, not every backup solution actually makes use of this bit - I know mine certainly doesn't. Reconciling firewall logs with server logs. Trivial.I beg to differ. Most packages designed for analysing and reconciling logs are incapable of applying a time shift to specific logs. Even worse, the PressF1 clock is not off by a fixed amount - the error constantly changes throughout the log. Name ANY package that can deal with a shift like that and I'll be impressed. And the firewall probably has the same time. as the web site server.What possible justification could you have for this statement? The only way they could be the same is if they were both syncing off an internal timeserver - and noting how the server clock constantly changes, I highly doubt that is the case. Most people who bother to set up a time service make sure that it's correct, as it tends to annoy people if it's not (and it screws some things up). Far more likely is the server clock has simply been left to its own devices independantly of anything else. Some types of server clustering ... all clustered boxes will have the same time. Who would not do that? The absolute cluster time is irrelevant ...The only way they would all have the same time is if they were using the same time service - which is likely to be correct for the reasons given above. I was not referring to the overall cluster time, but the interaction of hosts within the cluster. This particular point is irrelevant anyway, because as far as I'm aware PressF1 runs from a single server. Kerberos authorisation will not be bothered ... the K server will be in the same LAN, with the same time.Rubbish. By default, kerberos will refuse to authenticate anything with a time difference of more than a few minutes. Just because the KDC is on the same network doesn't automatically mean it has the same time. As for cookies, don't cookies contain the server's time as a data item inside the packet? The server is passed the cookie by the browser. I doubt very strongly whether the server has access to the user file system timestamp at all. That could be a major security problem. The user system timestamp on the file should be used only for the user system to remove cookies after some time. That's how I would programme cookies. I wouldn't be stupid enough to do that sort of thing in a way that relies on the time setting on thousands (or millions) of other computers (which I have no control over). The server knows its own time "now". It can compare that with the timestamp from a cookie it has just been given. Asking for the user's current time would be stupid. The server has better things to do. Also, as I pointed out, there are a hell of a lot of computers showing the wrong time. Many will be days, months, years out.This isn't my argument, it's vinref's. Many (on local networks which get their time set from the local server)...The only instance when computers on the LAN get their time from a local server is if someone has actually bothered to set up and configure such a system - which most of the time is not the case. The time on this forum can be a long way out causing no problems except for those haven't got watches, or who can't see the sun.I disagree. If nothing else, it causes tension between forum members and serious headaches for the resident pedants - would you call that, and all the above, 'no problems'? I'm not convinced.I am. Have I won my argument yet? |
Erayd (23) | ||
| 569784 | 2007-07-19 06:22:00 | just to stir the pot . . . . . . I am currently working on a project that involves setting up 20 server to get there time by NTP from the 5 local routers . the 5 local routers will get the time by NTP from our master time server . the master time server will be set about 10 weeks ahead :-) end result is we have a test LAN running 10 weeks ahead of the "real" world . as long as all server are wrong to the same amount the all is good . all so we can test daylight savings patches :-) as for the forum clock . . . . . I quite like it being a few minutes fast :-) |
robsonde (120) | ||
| 569785 | 2007-07-19 07:15:00 | This is great, in fact to keep my entertainment level up, we might just even make the clock even more wrong.Welcome to the PF1 ringside entertainment :D. There are a couple of thread subject matters that come with a 100% money back guarantee to always wind up the forum members - this subject is just one of them. :p You will undoubtedly get to see the rest over time ... |
Jen (38) | ||
| 569786 | 2007-07-19 07:20:00 | Another one . Bletch: I've been discussing the time "discrepancy" on this server . This isn't my argument, it's vinref's . I thought that all the things I addressed were vinref's . Show me the Kerberos authentication server on this forum . Show me the failed backups . Show me the problems with server logs . Show me the failed updates (where is the tar file with a date "in the future" going to come from?) . vinref introduced a number of problems which he seems to believe are inevitable if the time is wrong on this server . Computer networks work . They work together in internets . Internets work together in the Internet . They even work if individual computers have different times . The only instance when computers on the LAN get their time from a local server is if someone has actually bothered to set up and configure such a system - which most of the time is not the case . I was just mentioning this case from personal experience: The computers in the Public Library system here presumably get their time from the central server (since they all download their OS from the server) . They normally show the time as one hour "wrong" after the daylight time changes . It stays like that until someone gets around to changing the setting on the server . IT DOESN'T MATTER . They all access the Internet . No problem . . . . This particular point is irrelevant anyway, because as far as I'm aware PressF1 runs from a single server . Blame vinref for that . He said it's a problem . I say it's not on this server . . . . By default, kerberos will refuse to authenticate anything with a time difference of more than a few minutes . Just because the KDC is on the same network doesn't automatically mean it has the same time . Rubbish, yourself . What Kerberos server on this system? Do I have to say it again? If I was installing a K server in a LAN , I would make sure that the times in my cluster/LAN/whatever were consistent even if the LAN time seemed "wrong" to people outside . And I wouldn't necessarily have an automatic synchronisation with an outside time server in a time critical application . Sometimes they have problems . (I believe the Microsoft one was giving random times for a while) . Sometimes there are good reasons for having a "local time" which does not coincide with the official local time . I've kept wall clocks on UTC, and computers for the same reason . If your backup programmes can't cope with a computer which has a time "discrepancy" it needs some serious work . It's easy enough to back up any files which have been changed or generated since the last backup . The MS Archived bit is one way to ensure that . It's easy enough in Unix . I mentioned one time related problem I have seen on this server -- the odd sorting when the time was set back, so there was a doubled "time period" . Perhaps I should have suggested sorting on the sequence number, rather than the time . That was an aside . Sometimes the PressF1 server has the wrong time . It has been wrong for some time, and the world hasn't ended . A transformer has blown up in New York, and an earthquake has demonstrated, again, the perfect safety of nuclear power in Japan . I think it would be stretching the "butterfly effect" blame those events on the flapping of a few PressF1ers' knickers because the clock is a few minutes out . |
Graham L (2) | ||
| 1 2 3 4 5 | |||||