Brian Lee wrote:
Nope, this guy was from San Diego. He had many other pearls of wisdom.
My favorite was when he designed a system that would dynamically load
an EJB bean implementation at run-time. He wasn't using delegates or
proxies or anything coded, he wanted to use deployment descriptors.
That was a fun one to peer review.
*shudder* Here's my current horror story.
Here, programmers (who shall remain nameless) implemented "custom tags"
in Java on their own: a text file is parsed looking for a single tag
prefix and appropriate code is run.
Of course, the file is parsed every time its accessed (at least the file
content is cached), generally by the "controller" JSP. At least it's a
general-purpose templating mechanism. Oh, no, it's tied to web apps
only, never mind. Oh, and you can't nest iteration or logic tags inside
each other.
The controller JSP is a 400-line if/then/else state machine that has, at
a minimum, five ways to direct itself to the next state, and methods to
implement the state. But at least those methods are JavaDocced. Oh, no,
those methods are in included JSP files, so I can't trivially run
JavaDoc on them. Dang.
Well, at least the fake custom tags can be replaced by JSTL, because the
database classes (also hand-rolled) implement some sort of Map
interface. Oh, no, they just provide a getString method. Some of them.
Others only provide indexed access, so you have to get the index
corresponding to the keyword first.
At least I can trivially separate logic from view. Oh, no, essentially
every single method wants access to the request and the Writer. Oh, and
template files get their data from a Map context, not a normally-scoped
variable, so I have to re-do all that, too.
Well, at least it's implementing a small, relatively low-traffic site,
and there isn't a whole wad of stuff I have to look at to bring it up to
today's standards. Oh, no, it's a pretty gigantic site, fairly heavy
traffic, and there's about 5-10K lines of nearly undocumented code and
template files. Plus all the back-end libraries, for another 30K lines
of cramp-inducing code.
But at least I have complete control over where the project goes now, so
I'm free to use something like Struts, Spring, etc. Oh, no, the client
"doesn't like frameworks."
At least in my hiring interview they said I'd get to innovate, so I have
the time to explore new technologies. Oh, you mean AFTER work hours.
I _wish_ I could get my hands on the original developer for a peer
review, but I can't, because he's getting his doctorate at CMU, and I'm
told he was a brilliant programmer, so it must be true. (He also decided
to do a large project in ASP.NET (not using C#, mind you, but VB.NET)
because it would "save round-trips to the server", and every time
something needs to be changed I have to track down the SQL embedded in
the code-behind pages.)
Last I heard he wants to use the source for this... thing for
"educational purposes," to which I replied "Hey! Big Win! A great
example of how not to do stuff, and if anybody else steals the code and
uses it we'll have an automagic competitive advantage."
Oh goddess, somebody save me.
Hurry.
Dave "Dammit Jim, I'm a Lisp and SmallTalk programmer, not a... whatever
low-life I've sunk to" Newton
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
