On 8 Nov 2005, at 16:06, Scott Marlowe wrote:
> On Tue, 2005-11-08 at 09:45, Tino Wildenhain wrote:
>
>> Alex Stapleton schrieb:
>>
>>>
>>> On 8 Nov 2005, at 12:50, Tino Wildenhain wrote:
>>>
>>>
>>>> Evandro's mailing lists (Please, don't send personal messages
>>>> to this
>>>> address) schrieb:
>>>>
>>>>
>>>>> Hi guys,
>>>>> I would like to know if it is possible to have more than 1600
>>>>> columns on windows without recompiling postgres.
>>>>>
>>>>>
>>>>>
>>>> I would like to know who on earth needs 1600 columns and even
>>>> beyond?
>>>> Hint: you can have practically unlimited rows in your n:m table :-)
>>>>
>>>>
>>>
>>> Well this screams random arbitrary limit to me. Why does this limit
>>> exist? What ever happened to the holy 0,1,infinity triumvirate?
>>>
>>
>> I guess it eases implementation and there is no reason to go so high
>> on columns either. The limit could even be lower w/o and hurts but
>> 1600 seems skyrocket high enough (read unlimited :-)
>>
>
> I'd have to vote with Tino here. Why worry about an arbitrary
> limit you
> should never really be approaching anyway. If a table has more than
> several dozen columns, you've likely missed some important step of
> normalization. Once you near 100 columns, something is usually
> horribly
> wrong. I cannot imagine having a table that actually needed 1600 or
> more columns.
>
> And, Evandro, nothing is free. If someone went to the trouble of
> removing the limit of 1600, we'd probably pay in some other way, most
> likely with poor performance. There are other, far more important
> features to work on, I'd think.
>
Oh wait, PG is written in C isn't it. I guess fixed size things are a
bit easier to deal with. Pardon me then :)