Re: Initial run of ./bin/dbupdate fails.

[Art]

Bill,

I appreciate your post very much and this IS a worthy discussion!

Since you seem to think it belongs on-list, so be it; my motive is 
positive but I may say some things that some might find offensive, such as 
in one of my quotes you sent to the list. I do NOT intend to offend.

I also note that this is only cursorily related to Media Goblin, but it 
might be helpful. A bunch of what I share here belongs more in the ears of 
the Python people, but I don't know any of them!

I appreciate your at sharing with me important facts. I was not aware of 
the relative "educational depravity" you describe, and super-limited 
exposure, but it makes sense that what you say is true, at least on the 
broader scale. ... I have worked mostly "on the back end." My 
contributions to computing have mostly been in the database world, though 
I have done quite a few things on the "front end", including having proven 
a few things about the object oriented "front-end" world that stands 
majority (dogmatic) opinion on its head and proves those who stick to the 
dogma haven't a clue of how to code or use the tools properly.

Since I got outed, a bit of a digression on one of my remarks - if you 
want to skip it, look for END DIGRESSION. To wit:

When I said "kids today", I wasn't blaming the individuals, but the 
educational system. ...Back in 1995, I left the commercial database world 
to join Michael Stonebraker as a computer scientist at UC Berkeley - the 
guy who's teams invented Ingres, then Postgres and then quite a few 
others. (I worked for him at Ingres - that's how we met - and "I was 
there" at Berkeley when Paul Aoki led the push to open-source Postgres - 
it was his gutting of the QUEL language and completing / improving its SQL 
language abilities that prompted the name PostreSQL. And I've been 
involved with Postgres ever since. I've been urged to do some new commits 
to the code and hold the record for being the longest committer! But I 
just haven't the time.)

But back to "kids today"; one day a PhD student, asking for help on 
something, said - and I paraphrase some - "Oh, no, I don't know about 
that because it's ancient history - nothing in computer science over 5 
years old is worth bothering with!"

I damned near fell out of my chair. And this is a PhD student?! I opened 
his eyes and he went on to fantastic success - I'm not claiming credit for 
his success, only that he came around to the idea that learning history IS 
valuable.

END DIGRESSION

 ... "Packaging" is an extremely important part of computing history. But 
since Python is so important to Media Goblin, well...

First, I'm glad the GMG folks are not heaping on the dependencies as 
others have. And I'm glad you and Olivier Mehani are working on it. THANK 
YOU.

Given the importance of Python overall, the lessons the Python people 
haven't picked up on (yet? Or, at least I don't see it) is illustrated by 
the capabilities of the likes of C and Java and a host of other 
programming environments: You can pick up and bring with you what you need 
and cut the dependencies to an absolute minimum, AND you can include code 
from elsewhere, "on the fly." But you have the option of packaging it 
together. For example, I can package a specific JDBC driver with my 
application and the system manager of the deployed installation doesn't 
have to, for the most part, know or care.

If the licensing model is the right one (there are a LOT to pick through 
in the FOSS world), then you can build your package containing all that it 
needs in one go. ...These are not the resource poor days of the past where 
a few bytes of storage mattered greatly!

So, if Python - or if it were possible within Python already - had a 
packaging capability to harness together all the bits of an app so they, 
and only they, would be selected at run-time - from all the possible other 
versions also installed on a deployment system, that would do the trick.

> /me is on day 5 now - i had it fully packaged (aka: de-pipped
> and installed) on day 4 - though it probably does not work yet,
> because i have yet to package the javascript dependencies - i too
> will give up (again), unless i reach the bottom of the
> rabbit-hole soon

I know how you feel and would LOVE to carve out some time to try it!

Thanks again.

AND I appreciate your saying that it's not impossible for Python - your 
and Oliviers work (and others I'm sure) give me hope for this field. 
Otherwise it's a flipping disaster for deployment.

Another problem, by the way, is coders nearly never bother with 
installation issues - they far too often see it as someone else's problem. 
And this is harming the situation, as you implicitly note.

> but that gripe applies to nearly all webby things - you are not
> likely to find any which are easy to package - java-based
> websites are usually worse; because they tend to download all of
> their dependencies as binaries

I haven't noticed which proportions of badly done packages are based on 
what - most have been a substantial mix, in my experience. But 
java-script... well... Node.js is a GREAT idea that's finally happened. 
But it's not a panacea because, in my limited experience, it doesn't play 
nice with non-Java-based code.

Unfortunately, I have wasted a lot of time looking at such things and I 
think the fundamental cause is trying to create a stateful experience for 
users based on a stateless paradigm - the web servers. I complained about 
this from the very start, but it fell on deaf ears. ...IDK if that 
resonates with folks on this list, but it was a GIGANTIC mistake from the 
start of the World Wide Web and some of us saw it coming.

Thanks also for pointing out the issue of licensing.

> ignoring the labor of packaging all of the dependencies, the
> sheer number of dependencies also makes it a licensing
> rabbit-hole - just to illustrate, i also looked at peertube, and
> gave up hope on that one, immediately - it's terse dependency
> manifest alone, is nearly 10,000 LOC, without a single mention
> of any license
>
> https://raw.githubusercontent.com/Chocobozzz/PeerTube/develop/yarn.lock

Question: Do you think it's possible that the Python folks could compel 
their community - such as the folks who run package sites like pypi.org - 
to provide their FOSS packages in a license permits then to be put in 
someone else's package, such as GMG?

THAT is the kind of solution that's needed here - and, apparently, all 
over the web!

Regards, and best wishes to all,
Art

--
Art Richards
Art@TheTroyPress.com
http://TheTroyPress.com