As per my previous blog, Nigel presented 'Software As A Service' to the Sydney Architet Council gathering recently. Here is my 'raw' notes from the roundtable discussions:

• from the enterprise architecture perspective, the central IT can function as a 'provider of service', 'hoster of service', hence all the architectural issues applicable to a SaaS provider could be lessons learnt for the central IT as well (eg. minimise code to cut/test; development of monitoring facilities for tracking SLA etc)
• Integration cost with SaaS is non-trivial, especially with existing internal packaged applications
• security and authentication is non-trivial, single sign-on across organisation boundaries, ownership of this master authentication data are all hard challenges
• latency outside of your organisation - how do we evaluate and manage this risk? do we have statistics on SaaS projects that have gone wrong? and lessons learnt?
• federation solutions will have to be more prevalent to make this work
• SaaS CRM has had a bigger take up in comparison to others, could HR be the next thing? or ERP? Hosted email/calendaring?
• one financial service organisation in the roundtable shared with us the steps they took to offload a particular business capability/service they have packaged up and given to a "SaaS Provider" to be readied for global market. This particular business capability/service was prime to be chosen because of its generic functionality that is widely applicable to the rest of the industry, and does not require sensitive data to be stored/shared.
• What is companies like SAP doing in the SaaS space?
• The services in 'SaaS' are very course-grain right now, will these services in 'SaaS' become more 'finegrain' over time'? is this like 'SOA'? can the lessons learnt from SOA days be applied here?
• most SaaS vendors use 'interfaces' not 'Web Services'. Would this do? definitely calls for the need/use of 'Contract for Service'.
• What's the exit strategy for getting off a SaaS provider? what about the data? how/can I get it back? do we need to write the conditions into the contract up front?
• the SaaS data mining capability can be seen as 'risky' from the individual consumer perspective. While it can be a value add, the 'trend' mined from this needs to be only observed in mass volume, to avoid privacy issues.