You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@couchdb.apache.org by st...@shimaore.net on 2012/09/24 15:46:56 UTC
Follow-up on: What's up dev?
Hi,
I'm writing as an contributor on the wiki, CouchDB advocate, and
user.
I wanted to follow-up on the initial question about project energy,
and Noah's response, first to say that I'd be glad to help with things
such as documentation, use examples, etc.
Noah asked about "areas [people] give a shit about", so here's what I care
about (if there are any trolls just have mercy and ignore them, thanks):
- As a user (that is, a developer who uses CouchDB as a back-end
database and front-end in some cases), CouchDB is:
1. an API;
2. the API is what Futon tests;
3. as long as things are "fast-enough" I'm happy.
Most other things I can work around by writing code.
(For example: the replicator is muchly useful but not core --
replicate[1] might be a viable option --; and even though it is a very
good idea, I never used the _replicator database because I had coded
for that before it was there.)
I really don't care whether I use Apache CouchDB, or something
written in Erlang, or anything else -- as long as it conforms to the API.
... I use Apache CouchDB because I want the reference API.
For couchapps I rely on a modified node.couchapps.js & Sammy.js but
that's my playground, not the server's.
The API must be clearly documented, backward-compatible.
I'd like to see more index types (GeoCouch, longest-match) integrated.
I depend enough on the technology that I'd start my own version of the
server if things went really bad.
- As a system administrator I want to be able to deploy private,
distributed CouchDB simply and reliably. I want to distribute it over
multiple physical locations, multiple clouds if I can.
As a sysadmin I understand there are a lot of moving pieces, and the
most visible ones (view server) aren't necessarily in Erlang.
I care enough about the stability of my CouchDB installations to
package my own builds.
OTP has little practical impact because package managers just do "stop"
and "start" on upgrades.
... I use Apache CouchDB because there's little momentum behind
BigCouch, and I believed it was going to be integrated anyway.
- As an advocate I had to learn to answer "isn't it dead?" questions.
On the other hand I have fun when people mention binary protocols.
But overall I'm advocating for a technology, not a given
implementation. (And I'd love to have fellow developers prove me wrong,
obviously.)
S.
[1] https://github.com/mikeal/replicate
Re: Follow-up on: What's up dev?
Posted by Noah Slater <ns...@tumbolia.org>.
Thanks for the feedback! Useful stuff. I'll let other people reply to
specific points if they want to.
On Mon, Sep 24, 2012 at 2:46 PM, <st...@shimaore.net> wrote:
> Hi,
>
> I'm writing as an contributor on the wiki, CouchDB advocate, and
> user.
>
> I wanted to follow-up on the initial question about project energy,
> and Noah's response, first to say that I'd be glad to help with things
> such as documentation, use examples, etc.
>
> Noah asked about "areas [people] give a shit about", so here's what I care
> about (if there are any trolls just have mercy and ignore them, thanks):
>
> - As a user (that is, a developer who uses CouchDB as a back-end
> database and front-end in some cases), CouchDB is:
> 1. an API;
> 2. the API is what Futon tests;
> 3. as long as things are "fast-enough" I'm happy.
>
> Most other things I can work around by writing code.
> (For example: the replicator is muchly useful but not core --
> replicate[1] might be a viable option --; and even though it is a very
> good idea, I never used the _replicator database because I had coded
> for that before it was there.)
>
> I really don't care whether I use Apache CouchDB, or something
> written in Erlang, or anything else -- as long as it conforms to the API.
> ... I use Apache CouchDB because I want the reference API.
>
> For couchapps I rely on a modified node.couchapps.js & Sammy.js but
> that's my playground, not the server's.
>
> The API must be clearly documented, backward-compatible.
> I'd like to see more index types (GeoCouch, longest-match) integrated.
>
> I depend enough on the technology that I'd start my own version of the
> server if things went really bad.
>
> - As a system administrator I want to be able to deploy private,
> distributed CouchDB simply and reliably. I want to distribute it over
> multiple physical locations, multiple clouds if I can.
>
> As a sysadmin I understand there are a lot of moving pieces, and the
> most visible ones (view server) aren't necessarily in Erlang.
>
> I care enough about the stability of my CouchDB installations to
> package my own builds.
>
> OTP has little practical impact because package managers just do "stop"
> and "start" on upgrades.
>
> ... I use Apache CouchDB because there's little momentum behind
> BigCouch, and I believed it was going to be integrated anyway.
>
> - As an advocate I had to learn to answer "isn't it dead?" questions.
> On the other hand I have fun when people mention binary protocols.
> But overall I'm advocating for a technology, not a given
> implementation. (And I'd love to have fellow developers prove me wrong,
> obviously.)
>
> S.
>
> [1] https://github.com/mikeal/replicate
>
--
NS