[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [saf] RE: SAF meeting today - symptoms/prescription
This exercise highlights one of the key challenges we face with defining the problem domain of SAF. I don’t THINK that any of us envision SAF as being just a generic information exchange standard. Perhaps adoption and interest in the standard will not develop until we become MORE restrictive about what SAF is, rather than trying to make it more all-inclusive.
We’ve deliberately steered away from any such restriction in the past, but perhaps we need to limit SAF to the domain of “alerting about problems that may exist.” We can still be generous in our definition of “problem” – an under-optimized system, for example, could be considered a “problem.” Perhaps SAF is really just a standard for generic business process exception handling.
To take just the first row in the below table as an example:
Invoices: Do we really want to say that SAF can handle “primary” data packet transfers for well defined business processes such as invoicing? My concern here is that there are probably already established information exchange packets for sharing invoices. I don’t think that we could possibly make any inroads in to the exchange of invoices, and in fact, attempting to do so may actually weak the argument for SAF (we won’t be taken seriously).
So, in my mind, sending an invoice via SAF is clearly outside of the scope. However, raising concerns about an invoice may not be outside of the scope of symptoms. For example, the following may be a better example of something that would be within the domain of SAF (at least in my opinion):
“Outstanding invoices exceed credit limits of customer.”
So, sending an invoice within SAF would not be a good fit (this is a normal business flow).
Sending a notification that an invoice had been paid would not be a good fit (this is a normal business flow)
However, getting a symptom about a chargeback on a credit card (for example) and using that to trigger a prescription (turn off customer account) MAY be in the SAF domain (even this could arguably be a normal business flow though).
So the basic mindset is – if a system encounters a PROBLEM (or an indication of a potential problem) which it CAN NOT HANDLE within its domain, symptoms becomes the standard for REPORTING that problem to SOME OTHER ENTITY. Now the danger with this approach is that someone might think it is “just another event format.” Symptoms should perhaps be defined as the “subset of events that require mediating actions by an outside source.” Hence, an audit event (“User XYZ successfully logged in”) is NOT a symptom, even though it is an event.
Thoughts on this anyone? My fear is that we are becoming a giant vacuum cleaner for generic information flow scenarios – and I don’t think we will succeed if that is what we become.
Here are the results of my brainstorm (in table below), so far.
Will hopefully have more on the call today.
- Outreach review
- Cloud Profile – potential prescriptions & symptoms across different business functions
Most of the meeting will focus on the latter (Cloud Profile).
Please be prepared to discuss candidate symptoms (and prescriptions) emitted from different business functions. Pick your favorite business function and “have at it”!
Below is a list of business functions (stolen from a McDonald’s website). I’ve organized into two columns – Cloud Consumer & Cloud Provider.