Subj : Net Development
To : Jasen Betts
From : Jan Vermeulen
Date : Tue Jan 07 2003 12:32 pm
Quoting Jasen Betts on Sun 5 Jan 2003 22:16 to Jan Vermeulen:
jb> 04-Jan-03 Jan Vermeulen wrote to Scott Little
JV>> How can you be 100% sure that you will get the old data back when
JV>> generaing an SLF list from the XML data
jb> Is it desirable to get 100% the same data back anyway?
It is a must.
jb> things like the ordering of the flags or the method used to publish
jb> internet address aren't critical and some systems work better with
jb> one form and others with a different form.
If you're not strict to start with, you create a nursery for bugs.
mb>>> Of cource we have to see to it that everything we change also can
mb>>> be provided in a backward compatible format for the sysops.
JV>> Ok, the intention is there. But how sure can you be that not even
JV>> one byte will get lost or damaged in the operation
jb> One way is to prove the software and specification mathematically,
jb> but first a design is needed.
Ok.
JV>>>> How can you be 100% sure that you will get the old data back
JV>>>> when generaing an SLF list from the XML data?
sl>>> You can't, but that's dependant on the broken-ness of the input
sl>>> SLF. Theoretically, the SLF -> XML conversion will only extract
sl>>> "known good" data, leaving the rest as undecipherable nonsense
sl>>> which XML native programs will ignore, but will be restored when
sl>>> converted back to SLF.
JV>> This implies that an XML list generated from the nodelist at one
JV>> place will not yield the same nodelist somewhere else. I don't
JV>> like that
jb> why? as long as it contains the apropriate information does it
jb> matter,
Do not start to get lax even before you got the specs. You're sure not to
succeed if you do.
jb> when the extractor has to produce SLF for that nodeline it cant
jb> know how the line was originallt organised, but it can express the
jb> information in a sensible way.
