mwg@petrus.UUCP (Mark Garrett) (04/21/86)
++ > From: Stuart Friedberg <stuart> > Out-of-band (OOB) signalling is a partial solution to two problems: > 1) Indicating urgent conditions > 2) Providing synchronization markers on otherwise uniform streams > > What work (even blue-sky papers) has been done to generalize the > problem and its solution to something more than "bookmarkers" and > "flags"? > > Let me try to explain and motivate the question at the same time.... All this fancy theory about relating separate datastreams by using the synchronization codes sounds very nice and abstract (and may very well generate lots of research and PhD theses etc), but from an engineering point of view there are two problems. First is the problem of random errors. Out of band signalling makes it easy and quick to recover synchronization. Any gain you might get from making the sync codes more complex might be lost by the time you waste getting back in sync when you take a hit. Second is separation of control from data. Unless you can predict the volume and content of your datastream very well, you need a simple and reliable way of getting priortiy messages through in case of congestion. It seems you are trying to take redundancy out of the signalling (and use the redundant resource some other way). It is that very redundancy (and simplicity) that makes the lower layers of the communications channel reliable. -Mark Garrett Bellcore