jjb[m] has quit [Quit: You have been kicked for being idle]
TarekElsayed[m] has joined #mlpack
<rcurtin[m]>
I'm going through all the mlpack documentation in `doc/` as part of the header-only transition and preparation for 4.0. All of this stuff was written for Doxygen, but to me it seems like all these tutorials could just be converted to Markdown, and then simply referenced from the main README... this, combined with ensmallen-style docs for each top-level mlpack method, would mean we could remove doxygen entirely; what do people think?
<rcurtin[m]>
there's a bit more maintenance burden like this (since we have to keep the docs up to date manually), but arguably our current setup is nearly impossible for users to navigate and discover parts of mlpack
<zoq[m]>
I like the idea, doxygen can be difficult.
<shrit[m]>
please let us get rid of doxygen, and to be honest move all the doc out of mlpack code base
<shrit[m]>
you can create an independent doc repo that people can contribute too and we can link it to the website
<shrit[m]>
no need to ship doc and tutorial with the code, everyone will go online to read the doc
<shrit[m]>
this is my thoughts I do not know if these are our best option
<rcurtin[m]>
yeah, I'd like to keep the function documentation in the code for developers, but for user documentation, I think it should be separate Markdown that's at a higher level; so, e.g., `Adaboost` should be documented with a couple of trivial examples, a list of the template classes that can be used, some details on what it does and what methods it has, and then a link to the binding documentation (and the same for every other thing in `methods/`)
<rcurtin[m]>
then there will be no need for either doxygen or all the overhead of keeping it configured and running, custom stylesheets for the website, or whatever