Re: Jason is getting MARRIED!

From: Robert Ramey (ramey_at_[hidden])
Date: 2019-12-08 19:09:43

On 12/8/19 10:47 AM, Vinnie Falco via Boost wrote:
> On Sun, Dec 8, 2019 at 10:44 AM Robert Ramey via Boost
> <boost_at_[hidden]> wrote:
>> ...
> 
> This all makes sense*. Please answer this question:
> 
>> Should there be a JSON archive for Boost.Serialization?

If someone wants it.  sure.  but I don't recall anyone actually asking 
me for this.  Given the example of XML, it wouldn't be hard for people 
to make their own and perhaps people have with out advising me.  I'm 
aware that many have made their own special purpose archive classes - 
handling complex floating point, versions which avoid using the file 
system and copy to memory for extra speed, etc.  And of course one can 
pipe the output/input of any archive class through boost iostreams for 
encryption, compression etc.  The boost serialization library is not a 
library of archive classes, it's a library to build archive classes and 
it includes a few examples.

> If your answer to the question above is yes then please answer this question:
> 
> What is the benefit of a "JSON archive" that is not achieved with
> other archive types?

For my purposes - nothing.  The text version is just fine.  But I felt 
the same about XML and some people found a use for it.  (I actually use 
it to display my archive data when I have the read/write out of sync). 
I suppose there's someone who uses xslt to convert the XML archive into 
something displayable like PDF or html.  But I haven't had occasion to 
do this.  I do know that some people have edited the XML archive with 
some success.  One can't add / delete fields but altering the values in 
place would work and might be useful.  I once under took a project to 
make an archive class which would produce an editable form - a very 
interesting idea - but I lost interest in it.  Sooo  someone could find 
JSON useful for some special purposes.

My real problem comes when people believe that the serialization library 
which is very easy to use can/should be able to do what google protocol 
buffers does.  It's not possible because it's a different job.

B/S C++ data structures <-> byte stream

G/P/C GP C++ data structures <-> byte stream
                               ^
                               |
                               v
                 custom C++ or other language code

So if you want to explore how your JSON parser might be involved in an 
alternative to google protocol buffers, that might be an interesting 
idea - so you might find it interesting to take a look at google 
protocol buffers - which BTW is likely more popular than Boost 
Serialization any way.