Old lamps for new
Old technologies need re-inventing
Letters A couple of items in Reg Developer have prompted responses that continue a theme raised here before – that old software technologies that were developed years ago to solve perennial business management problems are still probably the best solution and should therefore be resuscitated - or perhaps unknowlingly re-invented.
For example, a welcome email arrived the other day from IBM distinguished engineer, Mark Cathcart (one of that company’s rocket scientists on virtualisation, architectures and design) in response to my piece on the possibilities of re-inventing batch processing.
He kindly pointed me at a press release that, I blush to admit, had escaped my notice. This revealed that, back in May this year, IBM had announced Batch-on-Grid – yes, good old batch processing technology adapted to run on Grid architectures, mixing old-style batch job queues with autonomic management systems to control and allocate the workloads.
In a comment to that piece, Del Merritt bemoaned the fact that most modern servers are only about as good as an old DEC VAX ever was. Given that HP can now run the VAX’s native O/S’s on the Itanium platform, perhaps it will re-invent that box, and give Itanium a new market as a helpful by-product.
In a similar vein, Deepak Vohra’s tutorial on Ruby on Rails has been generating plenty of positive interest, but it has also rattled the cages of one or two among our - how shall we say - more mature readers. And it is fair to say that the rattling has been caused by an arguably valid point.
As one correspondent put it:
'C' , C++, perl, Python, Java, Eifel, Ruby and Rails - those poor Unix-babies; they still don't get it?
The basic structures of 'C' and Unix are poorly suited to both business application development and device control.
Webservers are document presentation tools, not transaction processors.
Oracle is for data mining static tables, not real time posting.
One size does not fit all.
Quit trying to re-invent the wheel. Learn NS Guardian, System 360 (or whatever IBM calls it these days) Pathway CICS, DB2, NS SQL and COBOL.
It seems a fair point. There are technologies that are suitable for the glitz and gloss of presentation and the tricks unique to the web, but when it comes to what often lies behind that stuff – the need to run a business and not have it go belly up – most of the needed technology is there already, and has been for years.
Tell me I’m wrong, but it does seem like the software industry wants to invent new technology versions of "old lamps", when the old lamps are still better suited to the job. ®
SOA = Lipstick on Pig
Some of the big motivating factors for moving away from the old lamps is an inherent desire of developers to get closer to a new technologies capabilities thoughout a systems aspects (yes, I am referring to the interest in AOP even if I do not wish to actually use it) and thats a big deal. Adding layers of indirection and wrappers around legacy technology to make it work with the rest of the business is the issue at hand and the harder it is to do this, the bigger the financial pressure to move away from what would otherwise be a perfectly acceptable tool for the job.
Of course this was the promise of SOA, Service Oriented Architectures; wrapping legacy systems and allowing a fluid interop with new business platforms and user interface technologies. Of course we are just shuffling complexities from one place to another - managing a core technology migration instead becomes managing a zillion interfaces and service clients all of whom do different things. This is not to say that this has not been successful in many cases but it certainly is not the painless nirvana marketers would have us believe.
The really big deal about SOA and now virtualization is that it is thinking about the business not just the technology. There are some vast investments in technology and platforms and wrapping them up through services, virtualization or both is a good way of managing the shifting targets of performance, scalability and COST whilst allowing newer platforms to be a little more agile in their approach (Service orientation can be a good thing).
So what does this mean for the Old Lamps? Well, for sure it means that there is some longevity in the original technology and the irony of this is that there must be skilled people out there to maintain it. What do they do when they stop maintaining those old systems running in a virtualized VAX-11?
Write grid batch processors I'm sure...