From: Thompson, B. B. <BRY...@sa...> - 2006-02-03 18:10:31
|
Kevin, I am not convinced that the txid is required in the safe entry. -bryan -----Original Message----- From: jdb...@li... [mailto:jdb...@li...] On Behalf Of Kevin Day Sent: Friday, February 03, 2006 12:01 PM To: JDBM Developer listserv Subject: re[2]: [Jdbm-developer] transactionId : int or long? Bryan- I don't think time stamps are the way to go either. I'm confused about why you are writing any tx information into the safe... My understadning is that the safe should only contain comitted rows/pages... If data is in the safe, that implies that it was part of a tx that comitted, in which case having an idenftifier isn't going to help anything... Am I missing something important about the safe implementation? thanks, - K > Kevin, A txid is also one of the mechanisms which can be used by MVCC. See <http://en.wikipedia.org/wiki/Mvcc> http://en.wikipedia.org/wiki/Mvcc. Timestamps are less reliable since millisecond resolution is no longer enough to guarentee that distinct timestamps are used for two Txs and nano resolution is not guarenteed across platforms for Java 1.5. I've been writing the txid in safe entry header, but I am not convinced that it is absolutely required. It does give nice feedback since you can clearly see when the txid changes. If the lpc bit is not set on the last entry with a given txid, then that transaction was incomplete. However it may be that this informationis redundent. -bryan -----Original Message----- From: <mailto:jdb...@li...> jdb...@li... <mailto:jdb...@li...> [mailto:jdb...@li...] On Behalf Of Kevin Day Sent: Friday, February 03, 2006 10:34 AM To: JDBM Developer listserv Subject: re: [Jdbm-developer] transactionId : int or long? Bryan- Is a txid actually needed? An object representing a given tx will always exist in memory while the tx is active... This object would be used for keying any lookup tables, etc... For long tx, I think all we need is to ensure that the file containing the tx rollback info has a unique name, and there's easier ways of achieving that than using a counter for every single tx... What do you think? - K >Some MVCC schemes use sequential transaction numbers. Right now I have the txId modeled asan int. I am wondering if it makes sense to change this to a long, e.g., in order to avoid/defer the case when txIds are wrapped around? I choose an int because I thought that a new txid sequence would be generated each time the store was started, but if the a total order over all txids is desired then this should be changed. -bryan < ------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! <http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&dat=121642> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&dat=121642 _______________________________________________ Jdbm-developer mailing list <mailto:Jdb...@li...> Jdb...@li... <https://lists.sourceforge.net/lists/listinfo/jdbm-developer> https://lists.sourceforge.net/lists/listinfo/jdbm-developer < ------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&dat=121642 _______________________________________________ Jdbm-developer mailing list Jdb...@li... https://lists.sourceforge.net/lists/listinfo/jdbm-developer |