Re: Oracle vs. PostgreSQL - a comment

Hi Stefan,

> On 01. Jun, 2020, at 07:36, Stefan Knecht <knecht(dot)stefan(at)gmail(dot)com> wrote:
>
> Okay I'll bite.
>
> Comparing Postgres with Oracle is a bit like comparing a rubber duck you might buy your three year old, with a 300000 ton super tanker.

yes, and no. You are right about Oracle having gazillions of features but your comparison is way too drastic.

But be honest: How many features do you actually need? Most people use create table, view, sequence, index and that's basically it. Few use XML tables, Java inside the RDBMS, some (unfortunately) use Oracle Text. Many use BLOBs (instead of CLOBs) to mitigate the varchar2(4000) problem. Bottom line, most applications happily perform (even much better) on not so huge monsters.

> The rubber duck barely tells you how and why it floats, but the super tanker is packed with instrumentation, statistics, events and trace functionality down to every last bit of activity.

yes, but why do I need a huge hex block section in some trace file? Only Oracle can read that anyway. I don't have that with PostgreSQL because I don't need it.

And I am never sure if I deliver data to Oracle if I upload a trace file to them. Oracle support (sorry) sucks anyway. It's slow and in 99.9% doesn't solve the problem. I even abstain from opening service requests for years now. And my teammates still opening (and not having given up) service requests never get their first answer sooner than a day or two after the question even though the license says otherwise.

> Of course, that comes at a cost.

... excessive, that is...

> It's not a fair comparison.

I think it is because the user experience counts. It's like the iOS vs. Android religion. If iOS does exactly what I want then I don't see a need for thousands of tweaking features that Android (probably) has. Same with PostgreSQL. I don't need something like "alter session set events '10046 trace name context forever'" and learn that by heart. Why should I?

> Postgres has its place, it's free, it works well.

most definitely yes.

> But you can't compare it to an RDBMS like Oracle. Not in terms of size, nor the time it takes to install (and your 2 hours are definitely on the high end - it shouldn't take much more than half an hour).

I see that differently. As for the two hours: that is manual work just as is when installing PostgreSQL. Having done that once is enough of course and then it's packaged into Ansible for distribution. It's not about the 2 hours per se, it's about the big "much more" one has to do in any respect.

> In fact, you likely want to limit the feature set you are installing with Oracle

yes, I know chopt. Still...

> also to reduce the time it takes to install, upgrade and patch it. There are ways to do that.

yes, I know, which sometimes involves additional databases, storage and VMs, network, firewall rules and the whole nightmare which takes 4-8 weeks to implement because there are 4-5 departments involved.

Why not just limit the downtime as drastic as can *easily* be done with PostgreSQL in the first place without the whole setup nightmare that Oracle requires? I've been asking myself that for ages and always wondered why it couldn't be just as easy as it is with PostgreSQL.

Cheers,
Paul