Wednesday, August 3, 2016

[OT] Reuse Or Reimplement ?

I said in one of my earlier posts I'd share some thoughts I've had during the years on re-using vs re-implementing while working on various CXF projects. Some of it may be a bit CXF specific but most of it might be interest to non-CXF developers too.


When the time comes to implement a new feature the immediate decision that needs to be taken is how you do it. In general it is always a good idea to re-use a quality 3rd party library that can help in realizing the new feature fast.


Consider a task of providing a UI interface to have Swagger JSON documents nicely presented. One can invest the time and write UI scripts and pages. Or one can download a well-known Swagger UI module.

Another example: one needs a collection sort algorithm implementation which will do faster than Java Collections code. One can invest a time and write a new library or look around and try an Apache or Google library.

In both cases re-using the existing solution will most likely be better and help deliver the higher-level, complete offering faster.

Things may get more complicated when one works on a project in a competitive space. For example, at some point there were 6 active JAX-RS Java implementation projects, with other non JAX-RS implementations such as the one offered by Spring adding up to the total number.

When you work on a project like that one a number of important decisions need to be made: how complete you'd like your project to be ? Is supporting HTTP verbs and reading and writing the data is all what is needed ? What sort of security support around the service you'd like to provide ? What other extensions should your project have ? How would you like your project be recognized - as a library or something bigger that offers all sort of relevant support for people writing the HTTP services ?

The higher the 'ambitions' of such a project the more likely the 're-implementing' becomes a viable option, nearly a necessity in some cases. In fact re-implementing is going all around at such projects.

I've been involved in a fair bit of re-implementation projects.

To start with we started implementing JAX-RS at a time when Jersey was already high. Why ? To have Apache CXF open to users with different preferences on how to do HTTP services. It was hard at times but it was really never simply because we wanted to prove we could do it.

The latest 're-implementation' was JOSE. Why ? I won't deny I was keen to work with the low-level security code closer, but overall, I wanted a CXF Security Story be more complete. Implementing it vs re-using the quality libraries I listed at the Wiki let us tune and re-work the implementation for it to be better integrated with the JAX-RS and Core security support so many times that it would be highly unlikely to happen if I were working with a 3rd party library.

I do not think re-implementing in an open way is not healthy. For example it has been acknowledged that having many JAX-RS implementations around helped to make JAX-RS more popular. Re-implementing may offer more options to users.

Or, reimplementing can prove a complete loss of time. Here are some basic 'guidelines' if you decide to try to re-implement in the Open Source:
- think not twice but many times before you try it
- if you feel the urge then do it, get the experience, make the mistakes, next time you will do the best choice
- never expect that once you re-implement something then everyone will stop using whatever they use and switch to what you have written - a lot of clever developers are working full time
- if you'd like others to use your project then you absolutely must love working with the users, don't even start if you think that it will be up to the Customer Support
- you need to have a support of your colleagues
- expect that the only 'remuneration' you will have is the non-stop work to keep the project constantly evolving

Yes, very often re-using may be the very best thing :-)

Enjoy, Happy Re-Using, Happy Re-Implementing :-)



 





No comments: