Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!uunet!snorkelwacker!apple!oracle!news From: danderso@oracle.com (David E. Anderson) Newsgroups: comp.databases Subject: re:variable length records in Oracle Message-ID: <1990Feb1.165256.20379@oracle.com> Date: 1 Feb 90 16:52:56 GMT Reply-To: danderso@oracle.com (David E. Anderson) Organization: Oracle Corporation, Belmont CA Lines: 82 In article , yang@presto.IG.COM (Young-Mee Yang) writes: >How about others, Oracle, Sybase, Ingres? Do they allow variable >record length? In what I consider an accurate response, but in hindsight, a risky light-hearted play on the 7UP tv commercial, I responded with my very first net message: Oh, good. Please show me your SQL statement that creates a table >|with a text column without a specified size. and: dsimpson@oracle (Dave) provides short explanation of oracle data storage and Larger Flame Thrower On: jkrueger@dgis.dtic.dla.mil (Jon) >Thanks for the clarification. As we see, Oracle has never had it, and >as far as we know, never will. >"It" is variable record length, not trailing whitespace compression. >The original article never asked about the latter, >the question was simply which databases support the former. >If you must put your marketing claims on the net, please limit >yourself to accurate information that answers the question put. >Thank you. >-- Jon dsimpson@oracle.com then responded with a detailed and very accurate explanation of how oracle handles size declarations and data storage. I would like to add a few comments: 1) In posting my original message, I made a mistake which caused Mary Winslow's name to appear in the header as well as my own. She is not involved in this discussion, she is the Vice-President of my division :-) In Jon's followup Mary's name appeared rather than mine; my apologies for the confusion. 2) I think a mountain is being made out of a molehill. Let's see if I can be more explicit. I am a member of the Oracle kernel technical support group. I answer customer questions every day. In the few years I have done this, users who have asked about data storage have always referred to fixed-length and variable-length records as those which pre-allocate space according to the data definition and those which dynamically allocate space as a function of actual data size. I do not think it is unreasonable or surprising to use the terms this way and I certainly do not think that it calls for impuning my motives or asking for my silence. Using these definitions, Oracle clearly supports variable-length records, not a radical claim (or for that matter, a very commercially competitive claim, as I assume that most of our competitors provide the same functionality). I would like to know what definitions are used by Jon and, more importantly, the original poster... 3) To get on the right track, I suggest we discuss how software works and how we want it to work, which is what the original poster wanted to know. Dave Simpson has already described how Oracle works. I would add only that the user-defined maximum data size for a column can be changed at a later date using the ALTER TABLE MODIFY command. I would also like to take this opportunity to thank those of you who carried on the interesting threads centered on clients/servers/tp-monitors and oodbs over the last few months. David My original message was short because I thought this was a straight-forward, unambigous topic; live and learn, especially about the dangers of humorous/provocative wording!