Bringing SAP to dock

IN 1994, the first general elections might have been held in South Africa, and Euro Disney might have had to stump up another #1.6bn in funding after discovering that Mickey Mouse wasn't such a big cheese outside of the US, but as vintage years go, this one was fairly lacklustre.

On a smaller scale, though, things were definitely stirring that year for a group of independent software vendors with a marked interest in the AS/400, who had finally decided to proposition the IT industry's own big cheddar - IBM - on how the newly emerging star in enterprise software, namely SAP, might be countered.

Unlike SAP, the acknowledged giant of the top-end integrated corporate package market, the vendors did not have a bottomless research and development budget. However, this group of independent software vendors (ISVs) saw that the future of applications programming was object oriented. Would IBM help them migrate their offerings to object-oriented programming technology by putting in place some of the building blocks? After all, their success would mean more AS/400 sales, and possibly more System 3/90 sales, so IBM could expect a reasonable return.

IBM saw the wisdom of the suggestion and the San Francisco project was born. Its first deliverable, a general ledger 'framework' which gives the ISVs, in IBM's estimate, about 40% of a finished object-oriented general ledger, was announced in July this year and released on 15 August. Work is well advanced on the next set of deliverables - the building blocks and framework for the accounts payable, accounts receivable, distribution and logistics modules.

The distribution and logistics modules will cover functions such as warehouse management and inventory control, and will include business objects relating to many of the processes one would expect from these modules. While no release date has been set, the modules are expected some time in 1998. Other modules further down the road will include human resources and anything else IBM's ISV partners can think of.

What does all this mean? Is IBM getting back into the serious end of the applications business? 'Certainly not,' says Joe Damassa, IBM's US director of marketing for the San Francisco project. 'IBM is retreating from being directly in the applications business. What we are concentrating on is developing the lower level components of object-based business modules, to which our partners can then add their own expertise.'

This last point goes to the heart of one of the biggest dilemmas that faces ISVs who seek to reduce their own R&D input by basing their next generation products on a common, pre-fabricated framework and a thousand or so pre-defined business objects. If they use the same kit that everyone else is using, how do they differentiate themselves?

IBM's answer is simple: the key to successful differentiation lies in focusing your effort where it counts. You don't do this by trying to design a better calendar object than IBM has already created. You do it by understanding your client's business processes and building the rest of the module, the client-facing half of it, better than anyone else.

Although IBM has been remarkably low-key when it has come to its San Francisco project, it is hard to over-estimate the project's significance as a whole in the medium to long term.

US consultant David Andrews has been following the project from its inception. As he points out in a paper quoted on the IBM San Francisco Web site: 'It would be a mistake to assume that the low level of publicity means that San Francisco is not especially important or that it will only impact a small corner of the information technology world. San Francisco has quickly become the largest and most advanced server-based software project built using Java? Anyone involved in creating or buying packaged business applications has a stake in how the Java story evolves,' he argues.

Damassa himself points out that the Gartner Group has cited San Francisco as the largest Java project on the go at the moment. In addition, the Seybold Consultancy, he notes, has claimed in a recent report that IBM's work at the lowest layer in the project (the middleware components that it has added to Java) has moved the language forward at least 18 months.

The project currently involves a team of some 150 programmers and has absorbed more than #40m. This is not a particularly large figure, either in terms of IBM's overall research budget, or in terms of what vendors like SAP have been compelled to spend on new generations of their massive, integrated business applications suites.

SSA, for example, has already spent $500m (#295m) developing its own object-oriented Enterprise Resource Planning (ERP) suite, as the company's US Director of Global Alliances, Daniel Dibner, points out. However, Dibner is adamant that IBM is 'serious about San Francisco.'

'IBM's initiative will help to move applications programming globally towards the adoption of universal standards, based on object-oriented science, Corba and the rigorous definition of business processes,' he claims. Dibner reckons that his company has gained considerably by maintaining a very close relationship with IBM.

'IBM has not gone about San Francisco in a manner that is likely to expose it to the risk of failure,' he says, adding that anyone interested in object-oriented and Java is going to need to develop a close relationship with IBM for the foreseeable future.

'San Francisco embodies the programming disciplines of the next millennium - this has to be the future since it gives programmers a way of going forward that they never had before,' he concludes.

UK software house QSP, which specialises in financial software for large companies, is seriously considering using San Francisco as its route forward. QSP's research manager Mike Hay says that the company has already sent two of its staff on a 10-day training course on the San Francisco tool-set. 'I see it as a way for us to move QSP down from targeting companies with turnover in the hundreds of millions, to targeting companies in the #10m to #100m range,' he says.

Hay stresses that the San Francisco project is well past the vapour-ware stage. 'Our people were very impressed by what they saw. The presentation they were given contained over 1,000 slides and they were looking at solid product. The key technology areas of objects and relationships, base business classes, distributing objects around a network, two phase commit, persistence of objects - all these things have already been put in place,' he notes. 'All the areas we had identified as things we would have to address in moving our software to object-oriented have already been built.'

Hays points out that the potential savings for QSP are huge. 'Having 40% of the general ledger tower is nice, but the big savings for us come from the fact that San Francisco gives us a very sophisticated object-oriented framework that we can use directly. Doing this ourselves would have taken several years and would have been very expensive,' he says.

The licensing approach being taken by IBM with respect to San Francisco is also very attractive, he points out. 'There is absolutely no up-front licensing cost. We will pay only a single-digit percentage royalty figure when we actually began shipping completed product to our clients,' he explains.

ERP specialist Marcam has just completed its own huge development project over the last five years, re-engineering its suite of applications to object oriented, using a team of some 200 C++ programmers. However, Marcam senior applications consultant David Tudor says that the company will be watching San Francisco very closely as an option for any future development work - on an HR suite, for example. In common with other 'best of breed' suppliers, Marcam is struggling to extend the breadth and depth of its product offering to stave off the onslaught from SAP.

'Many vendors will see the necessity of moving to object-oriented, and San Francisco will view them as a way of leap-frogging the competition and getting product to market quickly. However, everyone in this industry is suspicious of quick fixes and developers basing their whole future on San Francisco will need to think carefully about it,' he says.

One of the major drawbacks commonly raised against Java as a mainstream business applications development language is the lack of performance comparisons with C++. Damassa points out that IBM has assigned the same specialists responsible for performance-tuning its top-end AS/400 series to the task of cranking up Java's execution speed and scaleability.

'When we started there weren't even any serious benchmarks for Java performance. We took three SAP applications benchmarks as our guide and we are now pretty close to them. What we are telling our vendors is that by the time they get their first products to market, in mid-1998, we'll have the performance issue solved,' he says confidently.

Damassa points out that IBM is already working to put Java on the AS/400 platform and expects to migrate it to the Series 3/90 as well. 'We'll be bringing enough expertise and horsepower into play to make performance a non-issue,' he says.

No precedent Impact viewed with caution

Dennis Keeling, author of the recent Ovum report on the corporate package market, remains to be convinced that San Francisco will have a significant impact on the package vendor arena. 'I have seen a lot of announcements heralding initiatives like this from IBM over the years, and very few of them ever see the light of day,' he says. All the successful vendors have had to do everything themselves, he points out. 'The concept of large-scale, complex applications based on shared objects has no precedent. I have yet to see anything succeed that is written in object-orientated programming in the financials area, for example,' he cautions.

The craft IBM signs up the tool makers

Tools chosen by IBM for San Francisco include Borland's JBuilder and Rational Software Corporation's object model, Rational Rose, as well as product from Symantec. Nigel Brown, managing director of Borland UK stresses, however, that while Borland is delighted to have Jbuilder as the Java tool of choice for San Francisco, his company will continue to produce tools aimed at the SAP market as well (Borland has just launched Delphi for SAP). Ian Gavin, UK marketing manager for Rational, points out that companies wanting to take advantage of object-orientated programming will be dealing with thousands of business objects and will need a tool like Rational Rose to manage the levels of complexity that this entails.