The official example template of creating a blog with Bootstrap.
comments: true published: true title:The web is not a good fit really.. tags: web internet rethink publishedat: 24/12/2014
Mobile and web apps are in wild demand and built by and for everybody. Yet nobody knows quite yet how to consistently and reliable structure and organize the back nor the front end. Some solutions are more popular then others, but there is no consensus, and fashion and fads seem to have undue influence. I want to talk about why and how this has come about and some ideas on improving the situations.
Companies are realizing that the traditional back ends based on java or php are not the most flexible or nimble, dare I say agile, and are jumping either on the nodejs bandwagon, or trying alternatives such as a scala (play) or python (django) framework.
Frameworks, systems and libraries that were built to create and serve "documents" have been and are being adapted to serve "applications", or to use the more usual awful misnomer: "dynamic documents". The first step on this road was made when static file servers were modified to execute scripts instead of serving a static file (cgi) when receving requests, and logically led to servers that were a runtime itself (nodejs). From complete configurability (apache) to complete programmability (nodejs).
In the meantime web sites, web apps and mobile apps need to be served and supplied with data, sometimes on an enormous scale. Website hosting service providers are being replaced with providers of raw computing power and internet bandwidth (aws, gce). The problem of scalability and distributed computing and data is not solved, with companies either outsourcing the solution, or building proprietary systems, or a combination. Old ideas and technologies are being rediscovered and/or repackaged (docker, erlang) in an effort to get a handle on the problem. Companies that are quick to learn and adapt or are new (startups, think whatsapp) see their advantage and are able to significantly increase their chance to unsettle older and bigger companies.
Yet no matter how easy to use the tools on both back and front end are becoming, and how much the back end infrastructure and service providers are adapting and improving, a lot of practices are still embedded in the old paradigm of the web as a document serving mechanism.
Functionality needs to be delivered to people, better known as consumers if you're the money making kind. The old way of doing this was by building a desktop application, writing it to a magnetic disc, shrink wrapping and selling it over the counter. An application developer only had to deal with one computer, one (fixed) set and source of data (disc) and one user. The user interface was a 'view' of the 'data'. No mvc framework needed. No network latency or disconnection to worry about. No browser incompatibility to work around. No endless libraries and frameworks trying to solve the same problem over and over. Etc, etc.
Come the web app and the thing is cleft perfectly in two. Most of the programmer's work is spent on adapting, writing, working around limitations of both halves, working against a legacy that was never designed to deliver application type functionality. An application developer is now worrying about load balancing, distributed data, sharding, document naming (dns), paths and urls, asynchronicity, state management, the list is endless. In the end however all this is boilerplate. It unnecessarily complicates the designing and implementing of the raw functionality required for a particular application, as all boilerplate does.
Ideas are floating around to radically redesign the internet. Naming and looking up things could be vastly improved. Using a human readable text string to locate and uniquely identify resources seems very antiquated, and very inefficient. Documents have localized names so we can find them in folders in filing cabinets. Computers are better at naming things for their uniqueness than humans (hashes). This is an idea from Joe Armstrong (see video 31 minutes in). Peer to peer protocols (BitTorrent) and other distributed technologies such as blockchains (bitcoins) are ignoring the old server-client model and embracing networks for their intrinsic qualities instead of reducing them to a medium to transport data from A to B.
Academics involved with the original design of the web as we know it now are rethinking its basic underlying premises, and want to use magnet links instead of urls to locate and store content distributed in the network and not at a centralized server. Popularity of runtimes such as nodejs have promoted the idea of isomorphic web apps, using the same language on server and client, to the point of erasing the difference between both where any source code can and does run on both server and client (meteorjs), not necessarily easing the job of the developer, but conceptually certainly attractive. Other, but as of yet less popular solutions of the server-client schism propose to do away with the server side of affairs all together and deposit all functionality in the client (hoodie).
A clear solution is not yet in sight, but the problem seem big to me. In the next blog post I want to talk about and design an experimental working application development infrastructure that tries to challenge some of the more commonly held underlying assumptions of how to deploy a web application. Dns, load balancing, rest api, ajax, scaling and more are all under fire. If it seems plausible it'll be implemented..