Is Seaside still a valid option?

Question

Seaside just released a release candidate for the upcoming 3.0 version, so it appeared on my radar again. As I'm currently pondering what web framework to use for a future project, I wonder whether it's something to consider. Alas, most of the publicity for Seaside is from '07, which is probably one or two generations for the web. So I'm hoping that the community here can answer some questions

1. Continuation-based frameworks were pretty great when most of your workflow was mostly in HTML, e.g. form submits. For today's JavaScript-heavy environments, that hardly seems worthwhile anymore.

2. Is Squeak able to handle a reasonable workload? From other questions here and elsewhere, it seems that for proper scaling another implementation (Gemstone etc.) would probably fare better in the long run, but I don't have a proper idea how far away that is. Sessions seem to be rather expensive.

3. I know that comparisons are hard, but most of the articles you find on the net set Seaside and Rails side by side. How would combinations like Scala/Lift, Clojure/Compojure or Erlang/Nitrogen do instead?

Answer 1

I have answers to question one and two:

1. This is true. However since version 2.8 Seaside is not a strictly "continuation-based" framework anymore. Seaside uses continuations in the flow module only. Since Seaside 3.0 the flow module is even optional. Also note that Seaside has strong Javascript support since 2005, this is long before mainstream frameworks started to add Javascript functionality. Today Seaside comes with JQuery and JQueryUI support built-in.
2. Of course that depends on what you store within your session objects, but typically sessions are small (less than 20 KiB). Use the memory profiler in your application to determine the exact memory consumption.

Answer 2

And there is a new seaside book: http://book.seaside.st/book

Answer 3

I find the productivity of working in a Smalltalk IDE with a good set of abstractions outweights all other issues in engineering dominated projects. It works well as an enterprise system for a small company with about 100 (simultaneous, but not heavy) users on a single server (without going to SSD). Since 2007:

• Seaside has shown to be able to make the switch from html workflows to javascript ones;
• Seaside has been ported to a lot of different Smalltalks;
• Has seen Gemstone release GLASS;

The new 'cog' vm with much improved performance has been released a few weeks ago and shows great promise for improved performance.

Answer 4

Avi Bryant, the developer of Seaside, said that AJAX triumphs continuations in almost all situations. Nevertheless, you can build reasonably powerful applications with Seaside and AJAX, too.

The Application part of a Web-App can be done in other frameworks quite well using Ajax.

I think a Seaside integrated Smalltalk-to-Javascript Framework like Cappuccino-for-Clamato is missing, currently. I'd like to be able to build real Javascript-Apps using Smalltalk.

Answer 5

1. Javascript is awesome but being capable of dealing with complicated workflow in a clean cheap way in the server side (as Seaside allows you to) is preventing it to become obsolete. Economy and utility are things that gives return in the short and long run. But telling this in the abstract has no value at all. You should be talking about a precise application and deciding if seaside is part of your bunch of competitive advantages to form an equation that rocks (and knowing why).
2. About scaling workload with Seaside, the short answer is that you can scale it like hell yah (for the long answer check my answer here: Does Seaside scale?).
3. too unanswerable man :) rty a variation of what you're really trying to ask

I think the best thing you can do is a prototype of something in a weekend.

If you can do a prototype in two days and you can capture some attention and you enjoyed the developing experience of doing it with seaside then you'll have the foundation of your next thing.

It costs only your time (you can publish in an amazon server).

BTW, this week I've heard about a startup that made its prototype by hand (was everything static and they processed stuff manually). Pretty amazing and crazy and cheap. When they felt that they had enough traction on the idea (which the had) they implemented the app (with whatever tech, I'm sure is no challenge for a seaside developer)

Answer 6

In Smalltalk we have now three web frameworks to consider, besides Seaside also

• Aida/Web and
• Iliad.

Both later effectively solve three-like control flow, but without needing continuations. Both also have a very strong Ajax integration, actually you don't realize anymore that you are working with Ajax.

Both also scale in memory consumption well. 10.000 sessions spend 220MB in Aida/Web, that is about 23KB per session, which can be further optimized down to mere 400B per session. This means, that you can run not only but many websites from the single Smalltalk image. Of course you can always upgrade to load balancing solution, when you really need. Which is from my experience very rarely needed.

Comparing to Ruby on Rails, a friend of mine complained that he needs 50MB of memory initially for every webshop site he is selling. He then turned to the Aida/Web solution where he needs about the same MB for the image, but then just few KB for every additional webshop site.