Bug or feature?
in [Programming SQL]
|
Prev: Hash Match best?
Next: Optimizing queries with isnull
From: --CELKO-- on 31 Mar 2010 19:36 VARCHAR(MAX) is proprietary syntax so Microsoft can do anything they want in any release. Remember when BIT suddenly became NULL-able? I would recommend sticking to Standard SQL so you can port across products AND releases.
From: Tony Rogerson on 3 Apr 2010 03:02 > VARCHAR(MAX) is proprietary syntax so Microsoft can do anything they > want in any release. Remember when BIT suddenly became NULL-able? I > would recommend sticking to Standard SQL so you can port across > products AND releases. Can Standard SQL data type hold 2 billion characters? Please explain, also explain how that relates to SQL Server in terms of implementation. What do you mean "suddenly", it took 3 releases of "this is going to change" before they actually did it. This scare mongering between dialect and iso standard is quite frankly unprofessional. The reality is that ISO do actually suddenly drop features for instance the ordinal position on the ORDER BY. --ROGGIE-- "--CELKO--" <jcelko212(a)earthlink.net> wrote in message news:6703ecd1-6aa9-4a02-bdc5-3a4386412306(a)35g2000yqm.googlegroups.com... > VARCHAR(MAX) is proprietary syntax so Microsoft can do anything they > want in any release. Remember when BIT suddenly became NULL-able? I > would recommend sticking to Standard SQL so you can port across > products AND releases. > > > >
From: --CELKO-- on 3 Apr 2010 14:44 >> Can Standard SQL data type hold 2 billion characters? << Sure; data types lengths are implementation-defined in the Standards. I worked with one DB product that had unlimited integer precision. One of their show-off demonstrations was to encrypt data with a 1000 digit key. >> Please explain, also explain how that relates to SQL Server in terms of implementation. << MAX is a propriety token and not a numeric value; the standard requires a positive exact numeric. >> This scare mongering between dialect and iso standard is quite frankly unprofessional. << I always considered needless dialect for job security to be unprofessional. >> The reality is that ISO do actually suddenly drop features for instance the ordinal position on the ORDER BY. << The ISO process requires FIVE years between Standards and a long public review process with committees. If a feature is deprecated, then it is marked as such in the standard before the change. ORDER BY is part of a cursor so that ordinal position would make sense -- columns are identified by name, no position in a table. The <unsigned integer> was in SQL-92, hardly a sudden change.
From: Tony Rogerson on 3 Apr 2010 15:10 > Sure; data types lengths are implementation-defined in the Standards. > I worked with one DB product that had unlimited integer precision. One > of their show-off demonstrations was to encrypt data with a 1000 digit > key. Really and how did they define this unlimited integer position? Name the product so I can go and validate this is not yet another load of rubbish like the DB2 and deferred constraint rubbish you tried to pull off. > The ISO process requires FIVE years between Standards and a long > public review process with committees. If a feature is deprecated, > then it is marked as such in the standard before the change. And just how many years do you think 3 full releases of SQL Server equates to? A bit more than 5 years! > I always considered needless dialect for job security to be > unprofessional. SQL Server you can specify nvarchar(4000), varchar(8000) which limits you, you cannot declare varchar(21528232) for instance. So why limit yourself to 8000 characters? Just how are you supposed to stored a 12KByte worth of text data in SQL Server? I'm sure you would actually implement that as multiple columns like the daftness of your advice with passing in a CSV to a stored procedure. But let us look at your statement; a good professional (like myself) will try and keep to with the standard but with experience you find out where the standard is really bad and completely divorced from what business actually needs to get answers off the data held in the database. You only need to see the current trend away from Relational databases to realise just how bad the SQL language is; the NOSQL movement is very strong now and is growing pace. > ORDER BY is part of a cursor so that ordinal position would make sense > -- columns are identified by name, no position in a table. The > <unsigned integer> was in SQL-92, hardly a sudden change. I am not debating Relational Theory here; ORDER BY is part of the standard - they have removed the ability to use the ordinal position in the ORDER BY. But you made the point that the bit column change "suddenly happened" - when in fact it took longer to get that change in then it does for standard changes - I understand you won't accept that because it's flies in the face of your dialect V pure standard argument which frankly you always lose because "we" out in the real world realise the limitations of standard SQL. --ROGGIE-- "--CELKO--" <jcelko212(a)earthlink.net> wrote in message news:a6b77d36-d16c-4f3a-8702-8e6cb25a160a(a)z6g2000yqz.googlegroups.com... >>> Can Standard SQL data type hold 2 billion characters? << > > Sure; data types lengths are implementation-defined in the Standards. > I worked with one DB product that had unlimited integer precision. One > of their show-off demonstrations was to encrypt data with a 1000 digit > key. > >>> Please explain, also explain how that relates to SQL Server in terms of >>> implementation. << > > MAX is a propriety token and not a numeric value; the standard > requires a positive exact numeric. > >>> This scare mongering between dialect and iso standard is quite frankly >>> unprofessional. << > > I always considered needless dialect for job security to be > unprofessional. > >>> The reality is that ISO do actually suddenly drop features for instance >>> the ordinal position on the ORDER BY. << > > The ISO process requires FIVE years between Standards and a long > public review process with committees. If a feature is deprecated, > then it is marked as such in the standard before the change. > > ORDER BY is part of a cursor so that ordinal position would make sense > -- columns are identified by name, no position in a table. The > <unsigned integer> was in SQL-92, hardly a sudden change.
From: Kalen Delaney on 3 Apr 2010 17:11
Hi Joe You said: > I always considered needless dialect for job security to be > unprofessional. I absolutely agree with this. BUT, what about needFUL dialect for performance and usability reasons? -- HTH Kalen ---------------------------------------- Kalen Delaney SQL Server MVP www.SQLServerInternals.com "--CELKO--" <jcelko212(a)earthlink.net> wrote in message news:a6b77d36-d16c-4f3a-8702-8e6cb25a160a(a)z6g2000yqz.googlegroups.com... >>> Can Standard SQL data type hold 2 billion characters? << > > Sure; data types lengths are implementation-defined in the Standards. > I worked with one DB product that had unlimited integer precision. One > of their show-off demonstrations was to encrypt data with a 1000 digit > key. > >>> Please explain, also explain how that relates to SQL Server in terms of >>> implementation. << > > MAX is a propriety token and not a numeric value; the standard > requires a positive exact numeric. > >>> This scare mongering between dialect and iso standard is quite frankly >>> unprofessional. << > > I always considered needless dialect for job security to be > unprofessional. > >>> The reality is that ISO do actually suddenly drop features for instance >>> the ordinal position on the ORDER BY. << > > The ISO process requires FIVE years between Standards and a long > public review process with committees. If a feature is deprecated, > then it is marked as such in the standard before the change. > > ORDER BY is part of a cursor so that ordinal position would make sense > -- columns are identified by name, no position in a table. The > <unsigned integer> was in SQL-92, hardly a sudden change. |