[GT] Re: Re: Re: baktest_multi.pl old xml format

[Thomas Weigert <weigert AT mst.edu>]

RAS, as you know, this change has been made long time ago (I don't quite
remember but probably more than a year). There have been no complaints
about the removal of the XML interface.

I believe that it is easy for experienced user to work around such a
change by renaming and continuing to use the old script.

However, for new users it is totally impossible if every script has a
different set of parameters and ways to interact. You know how much
people struggle with defining systems. Now tell them that for this
script they have to define their system in a totally different way...
That is a complete non-starter.

While I usually agree with your conservative attitude towards changes,
this is one where the conservative attitude becomes a barrier for
acquiring new users and it really does not protect the old users in a
meaningful way.

All that said, we should now leverage the familiarity of
fr.gamberale
AT
tiscali.it (sorry missed the name somewhere) with the XML
interface to maybe include this interface in all scripts, if it turns
out to be useful.

We finally have a consistent interface to all scripts. I am totally
against making it inconsistent again!

Th.

Robert A. Schmied wrote:
> Thomas Weigert wrote:
>> The backtest files where made consistent in their arguments. So please,
>> if you find that there had been some mistakes in this convergence,
>> highlight these and we shall fix it. But not by converting to the old
>> version but to an improved version that remains consistent with the
>> other scripts.
>>
>
> the conservative among us might endorse an alternative methodology:
>   first do no harm
> and
>   secondly consider each script a standalone application
>
> since this 'made consistent in their arguments' effort has been shown
> in the past to be defective, primarily due to unintended consequences,
> the change should be un-done by renaming the current 'improved'
> backtest_multi.pl
> to something other than backtest_multi.pl and the old original version
> restored as backtest_multi.pl. then, if users of the original
> backtest_multi.pl
> want to embrace any of the improvements or make additional enhancements
> that's fine, but that should be done on a consensus basis, grounded in an
> adequately exposed and discussed testing methodology that shows the api
> is unaffected or demonstrates the changes for all to consider. finally,
> pod should be included and be consistent with the applications'
> operation.
>
> but readers do note that i completely endorse the principle that
>
>   gt applications should be consistent in their arguments
>
> however, it should be applied only when and where applicable. most
> importantly it shouldn't change the user api of an *existing* script.
> if it does and the author cannot make it backwards compatible with the
> original api then the application should get an alternate name.
>
> the passage of time will demonstrate which of the two versions is the
> most useful.
>
> ras
>
>
>
>> There is no usage of an xml format anywhere else, so we should avoid
>> that or provide it as an alternative everywhere...
>>
>> Th.
>>
>> fr.gamberale
AT
tiscali.it wrote:
>>
>>> hi again,
>>>
>>> as i wrote in a previous post im still using an old version of GT.
>>> Reading the history of mailing list i noticed that the
>>> backtest_multi.pl was
>>> changed because nobody know the format of xml file. I use the old
>>> backtest_multi.pl quite a lot and can give examples of the xml file
>>> format.
>>>
>>> I can use the new one without problems but the new format doesnt fit
>>> to the
>>> backtest_multi.pl logic: in backtest_multi.pl you need only one
>>> MoneyManagement for all the systems, in
>>> the old format you had a xml fileld different from the system
>>> definition, in
>>> the new one you have to define MM only on the last system line,
>>> otherwise
>>> backtest_multi.pl gives an error.
>>>
>>> before i change all my other programs that use the xml format id
>>> like to know
>>> if what i wrote will make restore the previous version of
>>> backtest_multi.pl or
>>> is not enough important.
>>>
>>>
>>>  
>>
>>
>>
>