Hi Thom,

  I suggest you don’t use recordmapper at all or DTLs neither are good for very wide csvs,  use a file reader and save the input to a quickstream. Pass the stream id to a custom process which walks the quickstream changes each line and saves to a new quick stream, pass final quickstream to outbound.  This will prevent sending individual lines to processes or Dtls (which if your csv is huge would be a massive overhead).  You can contact me and I’d be happy to help.

Patrick,.

   You do realise this post is 6 years old?, subscriptions is certainly not an immature FHIR technology, it wasn't 6 years ago and now is borderline an essential part of a a useful FHIR service and HealthShare/HealthConnet/HealthRandomWord doesn't support it still without a lot of custom work.

For those that care (?) in the end I had to write my own handlers using some undocumented hooks in the ISC code and a LOT of trail and error, plus some custom operations/processes etc.  It was a painful process to get to work (but eventually did).