I currently work on one of the most trafficked production Derby applications[0] and I must say that this is a very welcome change. Derby's creators have time and time again proven that they care about the nodejs ecosystem. By funding ShareJS's development, they're ensuring that an unbelievable tool (simplified OT) will be sticking around for quite some time. Writing code with Derby has been far and away the most pleasant development experience I have ever had.
If there's been one issue with Derby so far, it's been Racer. The data sync-ing engine leaks memory and is largely unstable. We've suffered far too much downtime due to these issues. Moving to a fresh backend with ShareJS enables Nate, Brian, Joseph and the rest of the team to rework the scope of the framework with some of the larger goals in mind – things like horizontal scaling and powerful conflict resolution. We have yet to upgrade to 0.5 (the upgrade guide was literally just released), but it appears to be a much more stable foundation on which to build a "real-time" web application. I'm sure we'll have to put up with bugs for quite a few weeks, but this is a powerful step towards reaching 1.0. Thanks guys!
That is a really cool site, and a really great idea! Nice work.
I'm curious why a "real-time" framework is needed for this type of app, though -- why not Backbone with REST? Is it so that you can view others' pages to see their habits/dailies/todos get updated in real-time...?
The gist of it seems to be that its a set of data models that make handling concurrency control and conflict resolution easier in the context of a single blob of data being edited by multiple users simultaneously (similar to Google/Apache Wave)
Is anyone aware if Racer uses OT to sync the model as a whole (OT on a tree structure), or if it is used on a per-leaf basis to merge operations on values? Firebase seems to have solved this issue (with their Firepad project) by using their tree/json sync engine to transmit the OT operation primitives.
OT is done at the document level. Documents are JSON objects of arbitrary complexity. Any mutation of the JSON object can be composed against any other mutation of the JSON object.
Firepad doesn't actually do OT of JSON. In Firepad, a string OT algorithm is used to merge text edits, and the metadata describing rich text is separately managed as span objects, which get modified in accordance with text changes.
Firebase provides conflict resolution of JSON structures using their Transaction primitive, which allows them to create a Software Transactional Memory used to update the journal that powers the text OT in Firepad. ShareJS inside of Derby uses a similar approach to allow multiple servers to access the journal, but it is done in Redis with Lua scripting.
I've been following both ShareJS and Derby because, while I believe that the future of web application development is in realtime 'holy grail' frameworks, I very much dislike many aspects of Meteor, such as disregarding npm, the lack of proper server-side rendering (without the overhead of DOM) and the lack of importance placed on conflict resolution between client writes.
I really want Derby, or something like it (eg AirBnB's rendr combined with realtime, OT models and collections) to succeed, however it really seems like Derby needs more community support and evangelism behind it, because I can't help but get the impression that most of the community focus in this space currently is on Meteor. As far as I am concerned, Meteor is the PHP of realtime 'holy grail' frameworks, as enjoys from just the same kind of popularity through 'worse is better' (where 'worse' is for the reasons I mentioned above). However, without a palpable, visible sense of community enthusiasm, a lot of fence-sitters such as myself are less likely to jump on board.
Also guys please do something about racerjs.com, it doesn't inspire confidence in the project when it is continuously down.
Am I the only one who would look up a possible product name before using it? There already is a (well known) software from ASF named 'Derby': http://db.apache.org/derby
Why would you start such a confusion? Why would you want to fight with that software for page ranking? Why does this mostly happen to JS technologies?
There are far more software project starts than good names out there. Finding a good name that is also not used by any other project can take a bunch of time and energy away from your code.
> Why would you want to fight with that software for page ranking?
The vast majority of software projects never make it far enough for the name to matter. You don't know which category your project will end up in when you start it.
> Why does this mostly happen to JS technologies?
Because JS is popular and makes it easy to start new ad-hoc code libraries. There are also a lot of half-baked libraries out there that will waste your time. So many developers end up making their own.
There are far more software project starts than good names out there.
I doubt that. Define 'good name'. Also, for open source software it does not matter that much what it is named. Who would care if Python was named e.g. Snake or Juice or Foo? As long as it is somewhat unique and catchy it should be fine, shouldn't it?
You don't know which category your project will end up in when you start it.
Exactly, you do not know it. If your product is successful you want it to be unique. Just imagine the number of wrong tags on SE. And if you product is not successful it is even worse to add unnecessary confusion.
Because JS is popular
Surprisingly and unfortunately, yes. But other technologies have had their time as well, e.g. I can't remember any ambiguous naming in Java. Anyone?
There are also a lot of half-baked libraries out there that will waste your time. So many developers end up making their own.
Ok, why not just name it jQuery then? Hardly an excuse for poor naming choice.
A good name is pronoucable, spellable, memorable. Common English words can be good choices. Compare the number of common English words (thousands) with the number of Sourceforge, Github, Google Code, Codeplex, Bitbucket, etc. projects (millions).
> Ok, why not just name it jQuery then?
jQuery is a) perhaps the world's most popular project in the domain of Javascript libraries, and b) not a proper English word.
'Derby' is a relatively common English word and two Derby projects are in mildly different domains. I don't personally know of anyone using either project (unlike of course 'jQuery').
> You don't know which category your project will end up in when you start it.
So true. I don't know how many times I've started building a native audio streaming app and all-of-a-sudden I realize I've really built a web-based real time strategy game.
Your points are valid... just thought that one statement was a bit absurd and quite humorous.
I think by "category" he means successful / unsuccessful. As in, not knowing if the project will make it anywhere when first starting it. That's how I understood it at least...
>The vast majority of software projects never make it far enough for the name to matter. You don't know which category your project will end up in when you start it.
The "categories" are:
1) Projects where the name will matter because the project is successful
and
2) Projects where the name won't matter because the project dies before reaching critical mass.
EDIT: apologies, this was supposed to be a reply to examancer
I appreciate the information, but my comment was entirely an off-topic jab at Scientology. Where applicable, I apologize for the lack of sarcasm indicators. :)
Meteor is more mature, but it doesn't do server side rendering or operational transform for document updates. As far as I know, meteor is also limited to a single server. Derby now supports scaling over multiple processes on multiple machines. On http://sink.derbyjs.com/ I get DOMContentLoaded (including all the page content) in 70ms.
This module is intended specially for spiders to be able to see something of what the rendered page would look like. That provides an SEO stop gap, but it does not have any effect on regular users. Server-side rendering the first request greatly decreases the first page load time, which is typically extremely slow for complex client-only web apps.
It's not really the same thing. Since Derby is natively rendering server side, it's blazingly fast. You get the best of both worlds, render the initial page load on the server, and then all subsequent loads are on the client (with the data being sent over the wire using websockets or browser channel). This means you're getting initial page loads in the 70-300 ms range whereas rendering via PhantomJS and then serving to the client is much slower and much more CPU intensive.
As far as I know, Meteor doesn't even serve those PhantomJS rendered pages to the average user - only to crawlers - presumably for the reasons above.
It's too bad that they didn't notice, or didn't bother to find out, that the name "Derby" has already been used for years by an open source relational database project: http://db.apache.org/derby/
While we've only written a MongoDB plugin so far, we have designed specifically to be work with any similar datastore. Contributions of other database plugins are welcome. The Mongo implementation is here for reference: https://github.com/share/livedb-mongo
However Sails wants to add OT's, and add-ons that do Meteor-style syncing. I know this is wishful thinking, but maybe Sails and Derby could find some way to merge, or otherwise work together?
Wait - I just looked at the SailsJS docs, and it looks like Sails is like Tower -- that is to say, a Rails-style framework with better WebSocket support (really, more like Merb+DataMapper, because of the easy json api + agnostic ORM).
But if that'd be the case, it's a very different thing from Derby/Meteor/SocketStream, where so much code is shared from client+server that the apps can often run offline and then sync data.
Yes, they have different goals - by design, Sails is frontend-agnostic. But we want to add a layer that supports the "template syncing" concept, however it will be an add-on on top of Sails core.
Aside from that, Sails brings a lot to the table (awesome ORM, asset pipeline, built-in API, etc). All we need are OTs and things'll be rockin :D
Oh I thought it already had OT -- I'm currently developing a system that desperately needs it, and considered switching the entire stack to nodejs because of it -- but at this point, I have to ship so I couldn't...
Can any one who's looked at event sourcing comment on the relationship between ES (which is basically a fold over a sequence of events onto a state object) and OT. Superficially it would seem that they are very similar, with OT perhaps being a way to combine the events before applying them to the state.
The reason I ask is that I'm thinking of the feasibility of modelling a disconnected client that continues working from some state. Could you think of this as collecting its list of edits and then applying them on top of the state of the world when that client re-connects (sort of like a git rebase)?
I've been working on a real-time collaborative app that does just that: the client collects a list of edits and tries to push them to the server. It always displays the state as the last known accepted state with the local edits replayed on top. The server accepts edits from clients, gives them a total order, and rebroadcasts them. It's intended to be used while connected, but the protocol has no problem dealing with arbitrary gaps in connectedness, so adding an official disconnected mode would be trivial.
Importantly, "edits" here are discrete and independent. They can conflict (in which case it just does last one wins), but they can't affect the meaning of other edits. This scheme couldn't be used for collaborative editing of a linear text document, since e.g. "insert 'a' at position 15" means something different depending on whether it comes before or after "delete positions 5-10".
That problem is what OT solves (by the transform function that lets you commute edits while preserving their meaning). So, yes, that kind of scheme is much simpler than OT, but less powerful. Whether you can use it depends on the semantics of your edits.
Congrats and keep pushing forward! Although I have really focused on Python in the recent months, I keep an eye on the Node landscape... Derby always looked promising, along with half a dozen or so frameworks as well.
O.k. I live in Derby and that is also slang for a blowjob in America and a horse race. Plus much more I don't care to mention because the others already did. Do you have to do this?
Derby is a slang for oral sex in the US? I've lived in the US all my life and have never heard that. Can't find that definition on urbandictionary either (at least not the first page of results)....
If there's been one issue with Derby so far, it's been Racer. The data sync-ing engine leaks memory and is largely unstable. We've suffered far too much downtime due to these issues. Moving to a fresh backend with ShareJS enables Nate, Brian, Joseph and the rest of the team to rework the scope of the framework with some of the larger goals in mind – things like horizontal scaling and powerful conflict resolution. We have yet to upgrade to 0.5 (the upgrade guide was literally just released), but it appears to be a much more stable foundation on which to build a "real-time" web application. I'm sure we'll have to put up with bugs for quite a few weeks, but this is a powerful step towards reaching 1.0. Thanks guys!
[0] https://habitrpg.com (source: http://github.com/lefnire/habitrpg)