home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.software-eng
- Path: sparky!uunet!gator!miles!rms
- From: rms@miles.com (Rob Schultz)
- Subject: Enough statistics, what do we do?
- Message-ID: <1992Nov18.215032.16720@miles.com>
- Organization: Miles Inc., Diagnostics Division, Elkhart, IN
- Date: Wed, 18 Nov 1992 21:50:32 GMT
- Lines: 75
-
-
- Ok, I have to admit that some of the statistics in my last posting
- were slightly biased. For example, the US produces far more military
- software than Japan. The amount of paperwork involved in US military
- software is positively astounding, and it drives productivity rates
- through the floor. Even so, the quality numbers and sheer volume of
- programming staffs indicate that there are several other countries
- actively seeking to dominate the software engineering field.
-
- It is my opinion that we need to work long and hard at improving
- our productivity and our quality standards.
-
- Front-loading the process is the first, and most obvious method of
- doing this. More time spent in defining the requirements of a
- product, more time spent in designing a system, can mean less time
- spent in coding it. In fact, through the use of highly structured
- designs and automatic code generators, we can almost completely
- eliminate the time spent coding certain types of software. This not
- only reduces the amount of time spent in development (and testing), it
- also completely eliminates certain classes of errors. Automated tools
- could be developed that would guide our software engineers through the
- design process and perform automated consistency checking at all levels,
- from requirements through low-level design, in much the same way that
- hardware engineers use CAD tools. Of course, once these tools are
- available, there is nothing to keep our foreign competitors from
- using them as well.
-
- We also need to provide all of our software engineers with extensive
- training in the most effective methods available for producing sofware,
- and provide them with the tools to make use of those methods. This
- is going to require a substantial investment on the part of our
- management, which may be very hard for them to swallow, especially
- given the short-term profit motive currently running rmapant in US
- industry. The tools required include both the software tools that
- will allow us to front-load our process to this extent, and also
- the hardware tools for those tools to run on. This means high-speed
- graphical workstations and high-speed networks, including a national
- data infrastructure such as ISDN and/or NREN. Antiquated methods of
- software development using 25x80 ascii terminals connected to a mini-
- or main-frame computer at 4800bps or 9600bps simply will not suffice
- for the software development needs of our future. (Yes, I still see
- these types of arrangements being used for software development.)
-
- Higher level languages that require fewer statements to accomplish
- the same functionality will go a long way to showing productivity
- gains in the programming phase of software development, even if auto-
- mated code generators are not used. Lower level languages should be
- used only at the lowest levels of hardware interfaces.
-
- Software re-use is going to become one of the most effective methods
- of reducing software cost and improving software quality. I have
- heard time and time again that software re-use is a fantasy, but I
- see those same nay-sayers re-using software every day. For example,
- most programmers I know use code fragments as templates for software
- with a similar purpose. Had that software been developed with re-use
- in mind, it could have been incorporated into the new package code,
- design, and test. How many of us use math libraries or operating system
- calls to accomplish things without having to re-code them? I see no
- technical reason for not designing software for re-use.
-
- Of course, there are management reasons for not doing these things.
- I repeatedly hear managers saying, "Not on *MY* project," or, "We
- don't have the time or manpower to design some other project's code."
- And yet, at the same time, I hear those same managers complaining that
- their software development is taking too long and has far too many bugs
- in it.
-
- Well, I guess I have ranted on long enough here, so how about those
- comments? I know you are all just waiting to start pounding on those
- keyboards.
-
- --
- Rob Schultz At Home: At work: +1 219 262 7206
- rms@andria.miles.com rms@miles.com
- {uunet|iuvax}!gator!miles!andria!rms {uunet|iuvax}!gator!miles!rms
-