- intro and informal definition: d/transfer
- encoding of transfer ancestry in XML:
see g[tSS] and g.a of gene.mod
( too involved with genetic code, to cleanly factor out as transfer.mod
- group interaction in systems of recombinant text is reduceable
to one-way gene transfers, and this is an architectural strength
- no other communication mode is required
- each transaction is one-way, requiring no co-ordination among users,
at any stage
- no possibility of contentious, competitive dialogue
(as ding-dong battles, in a Wiki)
- but still allows mutually beneficial dialogue
in multiple, disjoint transactions
- thus a communication medium that minimizes the *necessity* of communication,
and this we see as a probable design strength. Because:
- communication and its issues are the main problem
in almost all group work
- technology aside, it is the most likely cause of any failure,
even in face-to-face systems (human organizations)
- any design that reduces communication without hindering system function
will therefore strengthen it
- the biggest technological success in collaborative media
(CSCW, groupware, etc.) is, thus far, the Wiki
- it minimizes communications by removing
the necessity of coordination prior to, and during transactions
- cost of minimizing is exactly this,
the loss of co-ordination
- anybody can do as they like with the text, up front
- co-ordination follows later, enforced by the fact
that others have the same freedom to do and undo
- we go one step further, by removing the need for co-ordination
at *any* point
- apparent cost is loss of all coordination
- actually, only the single central text is lost
- breaking out into a population of multiple variant texts
- although co-ordination is optional for participants,
and may be deferred indefinitely, it still happens:
- often implicitly, or at a higher level of the system
- emergent effect of many, one-sided transactions
- privately selective, but conducted in public view
- multiplied over time and across the population
- arising spontaneously, or guided by invitation
- as in formation of group consensus
- better yet, consensi
Transferant is an abstract design variation (of plot, character, etc.), encoded as a gene complex. Not described yet, except a brief summary of the approach.
- transferand is a continuous region of text,
encoded as a sequence of leaf genes
- essentially copy and paste of text regions, but
* transfer is genetically aligned,
for ease of placement in the target context
* authorship remains traceable, for attribution
- implemented by a/u/transfer
- transferand is a continuous region of text encoded as single gene
/ a trivial case of region transfer
- detailed in ../approach-simplex-wide.xht.
- referring to recent transfer ancestry recorded by 'ancestor' elements
- per gene.mod
/ see also d/gene/note.xht#Maintenance-of-Ancestry-Records
- Normally, the ancestry record is updated with each transfer.
There are two types of update to consider.
The first reflects a transfer that causes whole sequence replacement.
A whole sequence (i.e. a gene) is replaced.
In this type of update, the ancestry record is normally copied along with the sequence.
At this time, the system compares the abstract form of the source sequence
to the abstract form of its own, prior transfer source
(‘tSS’ attribute of ‘g’ element).
If they mismatch, then a mutation (by the source author) is detected.
A new ancestor element is created to represent the mutant source sequence,
and it is added to the record of the target product,
such that any and all previous ancestors nest within it.
- to detect recordable mutations,
a special, conservative abstract form is used
- it preserves enough sequence content that we record, at the least,
all mutations that any ancestry trace (in future) might want
- The second type of update reflects a transfer
that causes subsequence replacement,
where only a part of the target sequence is affected
[either by replacement or addition or deletion (special case of replacement)].
In this case, the transfer source is a second parent to the target sequence that results
(the other parent being the original target sequence).
If the writer wishes to acknowledge the source,
or if the tool can reliably detect it on its own,
then a new record is appended to the target product's ancestry.
The record is prepared as described for whole sequence replacement.
- although records added, never deleted; except as below, depth trimming
[ region transfer commonly affects two levels of the gene tree
- a series that contains multiple whole sequences
(the usual case, as opposed to the trivial case of a single gene)
is itself a sequence, and may result in transfer not only of the genes,
but of their parent (or parents);
i.e. affecting not only the matching genes of the target
(by their whole sequence replacement)
but also their parent (or parents)
- this occurs only if, in replacing the target series,
the transfer happens to alter the parent sequence
- by deletion | insertion | rearrangement of its child genes (loci)
/ a typical serial transfer will have no effect on the parent,
if by ‘typical’ we mean replacement of child alleles,
all of the same loci
- if it occurs, each affected parent experiences:
| whole sequence replacement
- if the resulting sequence matches exactly the source sequence
- this may occur, though only a subset of its child genes are affected
- in this case, the gene is updated per whole sequence replacement
| subsequence replacement
- if some significant part of the target parent
matches the source parent, that did not match before
/ e.g. we might say that a new, single locus match is insignificant,
wheras double or more is significant (application dependent decision)
- in this case, the gene is updated per subsequence replacement
| neither
- if there is subsequence replacement,
but it is deemed insignificant
? but may affect even more, higher levels - FIX, clarify
- depth (nesting) must be limited, because this is an in-document record
- so only recent ancestry is recorded
(full is distributed across multiple revision lines)
- at any time, the depth of the ancestry record may be further trimmed
- being conscious that the trimmed record
is inherited as such during transfers,
and that excessive trimming would undermine the purpose of depth (robustness)
- trimming generally discounts any set of co-ancestors,
one of which repeats a descendant earlier in the record;
as repeats add no robustness
- records serve as components for the assembly of a full sequence ancestry
/ e.g. in resolveToSequence(sequenceRecord)
- the purpose of depth is to make the components more robust for this purpose
so assembly can survive the loss of intermediate components
- transfer ancestry (as recorded here) is only a rough sequence ancestry
- it is rough because it excludes some intermediate ancestor sequences
(hence the element name 'ancestor' instead of 'parent')
- specifically included are transfer sources
(hence the name 'transfer ancestry')
- specifically excluded are ancestors from which they were derived
by entirely original (transferless) mutation
(probably effected with a text editor) in the revision line
- these are not needed for a normal trace, for authorship purposes
/ they may be recorded in revision line revision histories,
but only for those authors who use a revision control system