Re: ZODB Talk Slides

On Tue, Oct 24, 2017 at 8:30 AM, Christopher Lozinski <[email protected]>
wrote:
...

> On Oct 24, 2017, at 3:04 PM, Jim Fulton <[email protected]> wrote:
>
> ...


>
> These talks will bring ZODB more name recognition, but I still do not
> think anyone will start using it because of this presentation.
>

You never know. :)


> I think there's too much focus on applications in the last part.
>
> Well the talk is about the ZODB Ecosystem.
>

Fair enough.


>  I think people want an ecosystem, not just a database.
>

But ZODB is trying to overcome connection to a framework.

You pointed to a talk video a few months ago that said ZODB was not tied to
Zope. Yay.

Then proceeded with a statement along the lines of "Zope, get over it."
Sigh.

...

> After my second talk they asked about indexes.  Indexes are so central to
>> relational databases.  So I added in a slide about repoze.catalog.
>>
>
> You should also mention Newt, which lets you index with Postgres.
>
> Newt is in the slide on ZODB storage options.  I think about page 22.
> Lower right hand corner.
> Maybe it should have its own slide.
>

I think it's especially relevant in the context of indexing though.
...

Thinking about it, Newt is huge in the market, because you can have your
> data in both places.  Can you update through PostgreSQL, or only
> through the ZODB?
>

Only through ZODB. The JSON representation is lossy, to make it more usable.


>
> And you should mention various tools that let you use external indexes.
>
> A URL or two would be helpful.
>

Hopefully others will chip in. I think a lot of integration with
ElasticSearch and SOLR are internal projects.

There's https://pypi.python.org/pypi/repoze.pgtextindex



> I wonder what the next missing question will be?  I wonder what it takes
>> to actually get users to use the ZODB?
>>
>
> IMO, the main selling point is ease of use. You touch on that in your
> slides. There are serveral facets of this:
>
> - transparent persistence
>
> Got it.
>

You already cover this.


>
>
> - transactionally-consistent caching
>
> Never occurred to me that this could be a problem.
>

https://martinfowler.com/bliki/TwoHardThings.html

"There are 2 hard problems in computer science: cache invalidation, naming
things, and off-by-1 errors."

> - Easy testing
>
> A URL would be helpful.  I have dug for this but with no luck.
>
> I need to write a topic on this for zodb.org.

There are 2 aspects of this:

   1. Because persistence is transparent, application code can be unit
   tested without a database.
   2. Integration testing that needs a database can use a possibly-layered
   in-memory database, via DemoStorage.

Testing most database applications is painful.

> - Garbage collection
>
> It is interesting, since I use it, I do not even think of garbage
> collection.  Is it a problem with other databases?
>

In the RDBMs world, referential integrity is a big deal.  This is automated
in ZODB.

Jim

-- 
Jim Fulton
http://jimfulton.info

-- 
You received this message because you are subscribed to the Google Groups "zodb" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
For more options, visit https://groups.google.com/d/optout.