On 11/27/13 10:10 AM, Simon St.Laurent wrote:
> On 11/26/13 7:47 PM, Uche Ogbuji wrote:
>> On Thu, Nov 21, 2013 at 1:28 AM, Michael Kay <
>> <mailto:>> wrote:
>> One thing that Jackson misses, though, is that there can be multiple
>> hierarchic views of the data: this doesn't apply to the indexed
>> sequential files he was originally working with, nor (directly) to
>> XML, but it does apply when you extend Jackson to other fields.
>>
>>
>> THIS! Just this! In the day job I spend a lot of time explaining to
>> people that too many mainstream systems try to organize information with
>> one hierarchy to rule them all, or even worse, they degenerate the one
>> hierarchy into a tabular form. In my experience, most real world bodies
>> of information have multiple possible hierarchical views, and an
>> enormous number of IT problems boil down to juggling the political
>> pressures of prioritizing one view of the data over another within an
>> organization, and with respect to different populations served by the
>> organization.
>
> Agreed - THIS!
>
> When organizations get past the initial XML culture shock issues (they
> don't always), this seems to be the largest long-term challenge. In
> many ways, it's what drives the quest for brittleness I've seen and
> heard about too many times.
>
> It also seems to be what drives some folks to RDF, where hierarchies
> receive less worship.
>
> Thanks,
Interesting ... seems to be a very broad generalization, and I think the
situation differs in different domains.
In my view for text (hypertext, whatever), hierarchies + some light
linking seems to be a pretty good model. For business process modeling,
geometrical objects, social relationship graphs, maybe not. At least
the links become more important than any hierarchy, which is secondary.
When you import / export / transfer in bulk, you are treating your data
*as a text* so the hierarchy is useful since at that moment you don't
particularly care about the relationships (they are not important for
the integrity of archiving/transport), you just want to treat your data
as a big hunk of stuff to be moved around, saved, etc, treated as a
giant unit. This idea breaks down somewhat with smaller messages since
relational integrity becomes impossible to maintain across
loosely-coupled systems without imposing rigid top-down constraints on
the evolution of the data.
But when you are computing (rendering 3d models, searching relationship
graphs, performing trend analysis etc), the relationships, indexes, etc:
things that cut across the hierarchy in multiple dimensions, are what
matters, not the rigid immutable structure that XML models well.
-Mike
