Hacker News new | ask | show | jobs
by varjag 5753 days ago
Not arguing with the rest of your post, but what would you write a large-scale DBMS with, these days? Surely not jpython?
2 comments

Perhaps with 2010 style C ;-)

But maybe, just maybe, I would first connect Oracle's prodigious moneyduct into a team writing a kick-ass Lisp compiler and then writing the best RDBMS using Lisp.

One can dream...

Not an RDBMS, but AllegoGraph is in Lisp and is the most high-end NoSQL out there. Literally, the Oracle of object stores.
If you have an Oracle to hand, have a look at TNSNAMES.ORA and LISTENER.ORA. Greenspun's law applies...
Fair point. I didn't intend to diss C, and if one were to write Oracle from scratch today, C is (sadly?) still probably the best option given its performance and portability constraints.

What I had in mind was that for anyone who is technically capable of maintaining the core Oracle DB code base, there are more interesting and rewarding things they could be doing instead.

I would use something like typed racket or haskell for all the critical parts to increase safety. Then I would check the performance bottlenecks and rewrite them in C if needed.

Then I would use a more programmer friendly language on top of that. Something like Scheme, Ruby, Python and the like.

Stratified design, basically.

> I would use something like typed racket or haskell for all the critical parts to increase safety

I am not sure about Haskell, but, from my limited experience, Erlang runs at levels very close to C (my C, at least)

In terms of speed? Erlang's not even in the ballpark that C is in. I'm pretty sure Haskell significantly outperforms Erlang as well.

http://shootout.alioth.debian.org/u32/which-programming-lang...

Well... At least my C is regarded as very readable ;-)