couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Garren Smith <gar...@apache.org>
Subject Re: CouchDB and future
Date Mon, 08 Jul 2019 12:21:16 GMT
CouchDB would be a pretty poor fit for running on a small embedded device.
But CouchDB's API is a great standard. I know of people that use PouchDB to
run on the embedded device and then replicate the data to a CouchDB server.
That is a pretty standard way of doing it and should support most IoT
situations. I guess in the future it would be nice to have a small
embeddable CouchDB like server that isn't written in Erlang or javascript
that could then replicate or communicate to a CouchDB server.

The other thing to look at is often replicating IoT data to a server is
overly complicated. Most the time the data on an embedded device is metrics
of some sort e.g speed, temperature or something similar. You don't need
replication for that. All you need to do is store that data locally until
you have internet access and then send the data to a CouchDB cluster for
processing.

On Mon, Jul 8, 2019 at 12:44 PM Chintan Mishra <chintan@rebhu.com> wrote:

> Hello,
> > There is no other database that does CouchDB-Style replication, that’s a
> very clear and long-standing unique feature. Our particular combination of
> JSON/HTTP/REST API is also pretty compelling. I’m sure we can make a better
> job of explaining all that to people and I’m happy to review PRs for our
> website, docs, and drafts for our blog and discuss details on the marketing@
> list.
> The ease of replication, specifically multi-master, is unmatched when
> compared to many other NoSQL DBs. But now that most NoSQL DBs have
> adopted JSON/HTTP/REST API we cannot count it as a distinguishing factor.
>
> Market changes with time and consumer demands. We are at Nokia stage of
> multiple industries mentioned earlier. They will achieve their
> iOS/Android stage. The closer we align ourselves to the emerging
> technologies the easier it will be for us to stay relevant in the
> upcoming changes. I still stand by my inclination towards making CouchDB
> relevant from Web to IoT.
> > If that was true, the ASF board would shut down the project, as it is
> one of the things we have to provide a report on on a quarterly basis. All
> future direction of the project is discussed here and for my part, I
> understand where the project is going, even if I’m personally not involved
> in all the initiatives.
> >
> > If anyone’s specific vision isn’t included in our roadmap, we can talk
> about that, but know that we have to be very careful about allocating
> project resources towards the things we know we can ship vs. the ones that
> are a lot of work for not a lot of gain that subsequently nobody signs up
> to actually do.
>
> CouchDB is a very powerful project even with an improperly communicated
> direction. And it solves 'something' better than any other DB. And pin
> pointing this 'something' is very challenging. This 'something' is the
> reason why people who use it fall in love with it. As a user this
> 'something' and the future direction both are unclear.
>
> Can you share a link to the mailing list where we can read about these
> updates? This will help us improve website, marketing material, and more.
>
> On 08/07/19 3:33 PM, Jan Lehnardt wrote:
> > Hi,
> >
> > While I agree that there are many things that can be improved with
> CouchDB (we have an issue tracker full of them e.g.), I want to address
> these two assertions:
> >
> >> lack of unique competitive advantage,
> > There is no other database that does CouchDB-Style replication, that’s a
> very clear and long-standing unique feature. Our particular combination of
> JSON/HTTP/REST API is also pretty compelling. I’m sure we can make a better
> job of explaining all that to people and I’m happy to review PRs for our
> website, docs, and drafts for our blog and discuss details on the marketing@
> list.
> >
> >> and unclear future direction
> > If that was true, the ASF board would shut down the project, as it is
> one of the things we have to provide a report on on a quarterly basis. All
> future direction of the project is discussed here and for my part, I
> understand where the project is going, even if I’m personally not involved
> in all the initiatives.
> >
> > If anyone’s specific vision isn’t included in our roadmap, we can talk
> about that, but know that we have to be very careful about allocating
> project resources towards the things we know we can ship vs. the ones that
> are a lot of work for not a lot of gain that subsequently nobody signs up
> to actually do.
> >
> > Best
> > Jan
> > —
> >
> >
> >> On 8. Jul 2019, at 11:16, Chintan Mishra <chintan@rebhu.com> wrote:
> >>
> >> Hello,
> >>
> >>
> >> On 08/07/19 1:32 PM, Robert Samuel Newson wrote:
> >>> Hi,
> >>>
> >>> I understand what marketing is, but I also understand how this project
> is organised. The dev@ mailing list is where project direction and
> development is discussed, and where all decisions must be made, which is
> why I asked that the conversation moved here. Not many are signed up to the
> marketing mailing list.
> >>>
> >>> I hear and share your concerns about “IBM employees” (and I am one,
> but was not one for the majority of my involvement with CouchDB, like all
> the others that are now IBM employees). We have bylaws and a code of
> conduct and a PMC with many non-IBM employees. It should go without saying
> that those committers and PMC members who are also IBM employees are mature
> adults capable of working in open source with integrity. Were that ever to
> stop being the case, there would be consequences.
> >>>
> >>> The conversations about future direction (foundationdb et al) happened
> here on this mailing list and everyone was welcome to raise the concerns
> you did. Those conversations, and RFC’s, will continue to happen here.
> Ultimately, the project moves forward in the direction that active
> contributions take it under the watchful eye of the PMC.
> >>>
> >>> Chitan’s thoughts on IoT are very welcome, which is why I wanted the
> conversation here, where project ideas can be discussed with everyone
> interested in them. I also hope my reply that pointed out why CouchDB is
> not yet ideal for IoT was constructive. It is not just a question of
> marketing CouchDB as it is today, it should inform CouchDB project
> direction.
> >>>
> >>> B.
> >>> either
> >> Johs raised some genuine concerns.We all are raising our voice/concern
> because we believe in CouchDB. I have failed to see the masses love it as I
> do. I have been using CouchDB for around 1.5 years, probably lot lesser
> than most other subscribers here. In this time, I quickly realized that
> CouchDB is sitting on an underutilized goldmine. I share the concerns
> raised by Johs regarding unclear market, lack of unique competitive
> advantage, and unclear future direction while keeping existing user base.
> These are all the things that confused me in 2016 while deciding a suitable
> NoSQL Document DB for my use case. After trying 3 NoSQL Document DB, I was
> lucky to finally choose CouchDB.
> >>
> >> Robert, I understand bylaws and other rules in place are for us to keep
> this civil, maintainable, and organized.
> >>
> >> We are living in an era where products are consumer centric and some
> discussions will not have a clear boundary. The discussions will revolve
> around tech and market. Most of the time we will end up discussing both. We
> will talk about arising the customer/developer/industry needs and how we
> can build CouchDB which helps them excel. I see why IoT stuff cannot be
> done right now. And I wouldn't recommend as it is today.
> >>
> >> CouchDB led the way by realizing that future infrastructure will hugely
> rely on HTTP and built the first DB with a HTTP API for accessing and
> manipulating data. Newer advancements in technologies will always demand us
> to rethink our market position. We need to discuss that few years from now
> when IoTs, spatial/location-based services(drones, self-driving vehicles),
> analytic, etc. are the new normal then who will still use CouchDB despite
> other products in the market. Once we have figured who we want to build
> for, we all can start focusing on building it.
> >>
> >> PS: If you have 180 seconds to kill: https://youtu.be/DFeeNLElBgU
> >>
> >> Regards
> >> Chintan
> >>>> On 8 Jul 2019, at 07:08, Johs Ensby <johs@b2w.com> wrote:
> >>>>
> >>>> I find Chitan's reflections very interesting,
> >>>> even if the current direction for CouchDB points in a different
> direction.
> >>>> The problem is that CouchDBs future is difficult to understand, as
> marketing discussions are discouraged.
> >>>> This is a very good example of how it is done:
> >>>>
> >>>>> Briefly, this mailing list is for marketing only. If you wish to
> discuss project direction, post to the dev mailing list instead.
> >>>> This reveils a total lack of basic understanding of what marketing is.
> >>>> Just as a hint: Marketing is not the same as promotion. Marketing is
> a concept and a dicipline that arose in the 1960's as two previous concepts
> failed to capture the challenge of bringing products to the market.
> >>>>
> >>>> The "producion concept" worked as long as the market was underserved
> and you could focus on production, the market would swallow all available
> products (early days of mass production and hitech dev)
> >>>> The "sales concept" was characterized by strategies for pushing the
> products to market.
> >>>> The "marketing concept" turned the whole thing around and focused on
> identifying needs and wants in the market to which products could be
> developed in order to create value for the user/customer.
> >>>> A definition of marketing used by my professor at Harvard Business
> School was: Marketing is to create value and retain a fair share of that
> value.
> >>>>
> >>>> The notion that writing blog posts is marketing, unfortunately brings
> us to the pre-1960's thinking of how to bring a product to market.
> >>>>
> >>>> I am increasingly worried about the CouchDB project being dominated
> by IBM employees, and that new versions will require major workovers of
> CouchDB without concern as to:
> >>>> - what market is targeted
> >>>> - what competitive position in this market is being targeted
> >>>> - how to build on the existing user base for future success
> >>>>
> >>>> The latter point is crucial in order to make any technology adoption
> work, this is based on another theory that has been known since the 60's
> ("The diffusion of innovation" by Everett Rogers), popularised and brought
> to hitech in the early 90's by Jeff More's "Crossing the chasm".
> >>>>
> >>>> Chitans reflections on how to serve markets in need of new solutions
> is the kind of marketing discussion that CouchDB would benefit greatly from
> inviting.
> >>>>
> >>>> johs
> >>>>
> >>>>> On 7 Jul 2019, at 19:55, Robert Samuel Newson <rnewson@apache.org>
> wrote:
> >>>>>
> >>>>> Hi,
> >>>>>
> >>>>> The main (current) difficulties in running CouchDB on embedded
> devices is running erlang there coupled with our clustering stack. To scale
> everything down to fit within the typical constraints of an embedded device
> requires effort, I don’t think couch “just works” there.
> >>>>>
> >>>>> The future release that replaces the clustering technology with
> FoundationDB is what I refer to by “getting harder in the future”. While
> fdb does scale down well, it might still need porting to platforms it’s not
> currently supported on.
> >>>>>
> >>>>> As for suitability for IoT, couch is certainly used in that field,
> but care must be taken, especially around the core notion that couch
> retains information about a deleted document forever. This small amount of
> data can build up, and you need to plan for it.
> >>>>>
> >>>>> B.
> >>>>>
> >>>>>
> >>>>>> On 7 Jul 2019, at 08:02, Chintan Mishra <chintan@rebhu.com>
wrote:
> >>>>>>
> >>>>>> Dear team,
> >>>>>>
> >>>>>> Quoting myself from <marketing@couchdb.apache.org>
> >>>>>>
> >>>>>> ----
> >>>>>>
> >>>>>>> 3. A *way forward for CouchDB is focusing on what it is
best*at
> viz.,
> >>>>>>> being a database for _*C*__luster __*O*__f __*U*__nreliable
> >>>>>>> __*C*__ommodity __*H*__ardware_. Being deploy-able at edge
devices.
> >>>>>>> Focusing on this will invite people building for IoTs towards
> >>>>>>> CouchDB. And this will drive a whole new set of users/customers
> >>>>>>> towards CouchDB and IBM's Cloudant project. We already use
> >>>>>>> Cloudant's sync-android and CDTDatastore in our startup's(Rebhu
> >>>>>>> Computing) product.
> >>>>>>>
> >>>>>> ----
> >>>>>>
> >>>>>> I would like to draw attention to this point. I firmly believe
that
> >>>>>> CouchDB can benefit by targeting IoT device developers. IoT
> >>>>>> developers don't want to worry about sending data from edge
devices
> >>>>>> to server for processing. CouchDB already has battle-tested
> >>>>>> replication strategy. Extending this for IoT devs will drive
the
> >>>>>> next age developers. CouchDB is already made for unreliable
systems
> >>>>>> and what is more unreliable than an IoT lying in the sea connected
> >>>>>> with worldwide GSM/CDMA (2G) network.
> >>>>>>
> >>>>>> Quoting Robert Newson's <rnewson@apache.org> response
> >>>>>>
> >>>>>> It’s a good point but not for the marketing list unless you
are
> talking just about producing blog posts or other materials to promote
> couchdb for that market?
> >>>>>>
> >>>>>> I think to make couchdb more useful for iot would require some
> development work. Running couchdb on embedded devices is already a
> challenge and it’s only going to get harder in future. As a server side
> hub, the retention of a small amount of data after document deletion
> presents further problems.
> >>>>>>
> >>>>>> Briefly, this mailing list is for marketing only. If you wish
to
> discuss project direction, post to the dev mailing list instead.
> >>>>>>
> >>>>>> Thanks.
> >>>>>>
> >>>>>> @Robert_Newson can you shed some light on the issues that arise
> while deploying CouchDB on embedded devices? Also, what is going to get
> harder in the future?
> >>>>>>
> >>>>>> --
> >>>>>> Chintan Mishra
> >>>>>> Founder and CEO
> >>>>>> Rebhu Computing
> >>>>>>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message