Re: An object-oriented network DBMS from relational DBMS point of view

On Mar 13, 11:05 pm, "Tony D" <tonyisyour..._at_netscape.net> wrote:
> On Mar 9, 11:44 am, "Dmitry Shuklin" <shuk..._at_bk.ru> wrote:
>
> [ lots of snip ]
>
> I have a personal rule that I try to stick to. It's quite simple: if
> it's in c.d.t., and has "object oriented" in the title, it's probably
> not worth reading. This rule does not testing from time to time, so I
> paddled through this thread. The following three statements of
> Dmitry's stand out :
>
> March 13, 6:04pm :
> "I don't warry about correctness." and
> "why? implementation of vtbl can be done extreamelly fast."
>
> March 13, 6:11pm :
> "Hm, without pointers I will be unable to solve my problems."

Yes this is a particular problem and will no doubt be the sticking point to any conversation here. The most fundamental lesson of modern database theory has been the realization that pointers aren't necessary to achieve anything where data manipulation is concerned.

Its also a particularly difficult change in perspective if one is using OO methodology, with its network approach, day in day out. One should not underestimate the wrench that this shift in mindsets requires (and that presupposes a willingness to make the attempt in the first place).

Nonetheless, a question that I would like to ask is whether OO requires pointers - are they inseparably ingrained in to the whole methodology or could some hypothetical programming approach still be considered OO even if it did away with OIDs?

>
> March 13, 6:15pm :
> "View contans not the same row, just a copy of attribute's values."
>
> Or, more succintly put, it's the classic programmer's question :
> "How quickly would you like your wrong answers ?"
>
> Ah well. Rule tested, proven still sound. (sigh)
>
> - Tony
Received on Wed Mar 14 2007 - 02:03:32 CET