I think that it is widely conceded that it is a good idea to use the most suitable tool for a given task. But what does that mean? There is a long list of conditions and factors that go into selecting tools, some reflecting immediate needs, some reflecting long term needs and strategy, and others reflecting the availability of resources, and these interact in many ways, many of them problematic.
I have given the genesis of Cherry Hillís tech evolution at the end of this missive. The short version is that we started focused on minimizing size and complexity while maximizing performance, and over time have moved to an approach that balances those agains building and maintenance cost along with human and infrastructure resource usage.
Among the lessons we have learned in the last few years, one of the most effective has been to employ services wherever possible. We have moved our infrastructure to AWS, use PAAS tools like Heroku and Joyent for development, and we use SAAS tools like Github, Pivotal Tracker,and Unfuddle for project management and testing, as well as business needs like telephone, meeting tools, and data management.
With regard to software, we employ the ďhire slow, fire fastĒ axiom that others apply to HR. Bad choices can be expensive.
When we adopted Drupal, we didnít dwell on the ďDrupal is hardĒ label that was a mantra of the time, as we did not find it to be hard. The parts we didnít like, we could fix.
In 2007, we started working with the library Drupal community to promote Drupal to libraries of all kinds, based on the virtue of strength in numbers. This extends the community principles that drive the Drupal project. With over 1,000 CMS systems in production, it would be surprising if one of them, no matter how flexible and adaptable was perfect for every situation. This is why, while we employ the Drupal brand and have an investment in our partnership with the Drupal Association, we prefer that prospective clients do not approach projects with a closed mind. If you ask us for a website, we are likely to propose it in Drupal, because we have a lot of tools and resources at our disposal, and we are facile in working with Drupal.
Apropos the original message, Drupal has great tools for supplying and consuming services, including the services and feeds modules. We are working on some projects that pair the Drupal CMS form the presentation layer. One of these came about when a client launched a major site rebuild (using a different contractor) with an Angular.js presentation layer on top of RoR. What became apparent pretty quickly is that there was no simple way to create, manage and organize the thousands of pages on this site in RoR. We were called in to create endpoints in the current Drupal system that would feed RoR, etc. This buys them time to build out a new, bespoke CMS in Rails.
Likewise, even though we have RoR projects on the books and we do a lot of work with Ruby in our system administration, we want to build new bespoke projects in Symfony, because we like it, and because I really enjoy it and because it is the future of Drupal. I love Python, Scala, Go and a bunch of languages, but we canít do everything. I donít like Perl, for reasons that I canít really articulate, and (I hope) you canít make me use it.
From a project ownerís perspective, I think that it is critical that you do your homework and work with tools you are comfortable with. This doesnít mean that you should build your social media project in Cobol. You should familiarize yourself with what is out there in the world of today, determine what resources are available, and divide what mix of internal and external teams you can work with. Bear in mind that the number one fail is having someone build a complex, wonderful, undocumented bespoke tool, which six months down the road, needs to be scrapped because it canít.
I believe in open-source, with a strong preference for FOSS project built by a community. I feel that it is a win when libraries build projects in Wordpress, SilverStripe, RoR, Plone or any open-source system. I donít feel so good about SiteCore, who loves to blast the idea of open-source, or SharePoint, which is, I m sure, good at something, just not at web delivery.
I am going to give you some insight into what my shop does, why, and how we made these choices. It is a snapshot, as we evolve constantly.
We have always been a multi-thechnology shop. Originally we were Microsoft partners, building desktop and distributed applications in Access, MSSQL, and VBA, which we augmented, mostly for speed, with com objects written in C++. We did our web development in ColdFusion, originally against MSSQL, and also augmented with com objects. When ColdFusion moved to Java, the com objects became Java, and we started migrating to Linux and MySQL.
We started using ColdFusion as an integration tool, and, when we started working on the 24/7 reference project, we used Java to connect to and communicate with the integration targets, using CORBA and other service bus tools. We also started working with WSDL, and other early service platforms.
Eventually, we started using ColdFusion frameworks, including FuseBox (meh), Model-Glue, and ColdBox, which in turn started moving to the MVC model. We were also doing some work with the Sakai LMS, which led us into the Spring and Hibernate Java frameworks. The big, overly simplified take away was that MVC frameworks made projects easier to manage and maintain, but there was often a performance hit. Ten years ago, servers had improved enough where we didnít have to do component sharpshooting to get applications to run, but they still did not have great performance. Apps that we now run in a VM or two, might require half a dozen web heads, substantial SANs and database servers, and expensive load-balancers.
In 2005, someone pointed out Drupal as something that might be a good fit for a no-budget project where we needed a useful fronted for dSpace.ColdFusion was not a option. Drupal 4.5 was imperfect, but it had a couple compelling qualities. Drupal core, which was pretty much all I used, was built with coding standards which elevated it above the craziness of PHP 3 and 4.The code also contained great embedded documentation. Even better, it was very XML friendly, which was key to getting the project done.
The timing was good, as we were about to rebuild our CMS service package for libraries ó LibrarySite ó and building that in Drupal would allow us to lower the price by about 70 percent.
In 2007, I went to a DrupalCon in Barcelona, because I was having some issues with a multilingual site project, and I know that the players in that initiative would be there. I solved my problem, and I was smitten by the community. I got involved, produced some DrupalCons, and served on the Drupal Association board for a few years.
Although most of our projects involve Drupal.we are not a pure Drupal shop. We are also doing some RoR projects, we touch Java through working with Solr and Jenkins, and Python through AWS scripting, and casual programing.
We are starting to do more with Symfony (and Silex), and we have a particular interest in JS-based systems including angular.js, node.js and more. One of our demo projects is using the mean.io framework integrating with Drupal. Mean.io is a mashup of MongoDB, express, Node.js and Angular.js.
Our first web applications were simple and flat. Then they were complex and flat. While flat applications might be performant, they become unmaintainable as they grow, and it is on us to find the balance between maintainability and structure, and performance and operational cost. All of this must be considered in an ever-changing universe of tools and techniques.
In the Ď80s, Microsoft Word was very popular, but lacked printer drivers. It did have a well documented system for created them that involved a hook-like arrangement. Consequently, when I first looked at Drupalís hook system, it was instantly familiar.