Aura HD A simplified High Performance AuraHD?

[Ken Maltby]

It is my impression that the complexity of the sync and tile based
user interface hobbles the performance of the AuraHD. It could be
much of the reason that the 1000mHz Aura can't handle the number
of files or a file structure of any real complexity, when compared
to ereaders using much less powerful 400mHz CPUs.

The power and potential of a database derived user interface may be
very alluring and the "book management" as opposed to file management
concept may work well for calibre, but a simple user manageable file
structure would allow for a higher performance with large ebook
libraries.

Just as when trying to make a high performance racing vehicle, a
stripping out of any extra, unnecessary weight, is an essential
step. So much that appears to be there solely to tie use of the
device to Kobo servers provides nothing to the basic reading
functionality of the device.

I know nothing of how the FW is structured but I do know that some
modularity must be there. Would it not be possible to unentangle
the reading and just those other components necessary to support
reading, from the current mix. Then remove the current overlay of
the file system and allow for the normal file management of the
partition's format.

I realize this approach would not be popular with Kobo, but it
would be just the thing for many of us users.

Luck;
Ken

Can you provide some examples of performance issues with the Aura, and I'll try looking at what's going on. (I can guarantee that I'll be able to pinpoint the problem, but I have just enough experience with Linux and its development tools that it should be a doable project.)

One thing that I can say is that a pure file system based solution wouldn't work in an ereader targeted at a general audience. Data such as the title, author, last position, etc. should be presented to those users rather than a list of directories and file names. Now that data has to be stored somewhere. While there are various ways to do this, a database is likely the most efficient means. Databases tend to be developed in such a way that searching, retrieving, storing, and removing data is done in an efficient manner. Using the data from the ebooks themselves would probably be the least efficient way to handle this, since it means opening and parsing data in multiple files when you want to present the data for a collection of books. (As added fun, some of that data may be compressed. Decompression adds even more overhead.)

There are other ways to handle it, such as a datafile per book (you're still accessing multiple files, but you can make the data more efficient to handle) or a datafile per directory (which really reduces to developing your own database, but it may be more efficient since it's application specific). Alas, many of those solutions make it much more difficult to add universal features such as searching your collection. That may not be an issue for you, but it almost certainly is an issue when the device has to reflect the needs of a general audience.

As for the tiles and sync hurting performance, I seriously hope not. Linux can certainly help background network operations without a noticeable impact on performance and the tiles shouldn't even be active in the background.

[ewphoenix123]

I think what comes closest to what you want is the Koreader app.
It supports file system based book selection/navigation and kills the other background processes afaik.

Problem is, the last version i tried was not working with the aura resolution (maybe fixed by now there is a newer version)

[Ken Maltby]

What's "a pure file system based solution"? All software stores data in a number of
ways to carry out program functions. There is nothing magical about an SQL database.

If a 400mHz jetbook Lite(JBL) can take seconds to build a menu out of the metadata
found in over 7000 ebook files, presumably generating its own index database in the
process, why can't the 1000mHz AuraHD function smoothly with a large number of files
and "shelves"?

You keep mentioning "the needs of a general audience", it certainly could be that the
"must meet all needs" approach is what has given us the current FW - but if you would
read my post carefully you might notice a theme based around the term "simplified".
There are some of us who do not want all the features you think can only be provided
with the current construction of the FW. Some, like myself, would prefer none of these
added features and have all the effort applied to a simpler file selection and reading
environment.

Kobo already made your bloated sometimes functional "general audience" ereader, I am
just asking if someone might consider slimming it down, a lot, and untethering it from
the Kobo servers.

Luck;
Ken

[Ken Maltby]

Originally Posted by ewphoenix123

I think what comes closest to what you want is the Koreader app.
It supports file system based book selection/navigation and kills the other background processes afaik.

Problem is, the last version i tried was not working with the aura resolution (maybe fixed by now there is a newer version)

Hmm... this does sound like they may get there along the way. I'll certainly
keep an eye on its development. Thanks, a lot!

Luck;
Ken

[PeterT]

Originally Posted by BWinmill

Can you provide some examples of performance issues with the Aura, and I'll try looking at what's going on. (I can guarantee that I'll be able to pinpoint the problem, but I have just enough experience with Linux and its development tools that it should be a doable project.)

What would be fascinating would be if someone could manage to trace the SQLite calls being made.

I'd like to see what I can do on the profiling end. I'm guessing it'll be limited to CPU, RAM, network, disk monitoring to see where the bottleneck is; but maybe I'll be able to go further.

Can someone help out by suggesting a large collection of ePubs that can be downloaded in one shot? Gutenberg is fine. Their site is less than clear on how to go about this though. (Text files, sure. DVDs with unspecified formats, sure. ePubs?)

[davidfor]

A good set would be those from the library here. The advantage of them of PG is that they will definitely be validated. I don't know if there is an easy download mechanism to get a lot of them.

[speakingtohe]

Originally Posted by BWinmill

I'd like to see what I can do on the profiling end. I'm guessing it'll be limited to CPU, RAM, network, disk monitoring to see where the bottleneck is; but maybe I'll be able to go further.

Can someone help out by suggesting a large collection of ePubs that can be downloaded in one shot? Gutenberg is fine. Their site is less than clear on how to go about this though. (Text files, sure. DVDs with unspecified formats, sure. ePubs?)

You could download the text files and convert them?

Helen

[Danger]

I just did a quick search and got this: http://www.gutenberg.org/

I then clicked on recent books and got this: http://www.gutenberg.org/ebooks/sear...3Drelease_date

Clicking on the first book gave me a list of download links to choose from, 2nd link was for an epub.

[Lucas Malor]

@Ken Maltby: does your Aura slow down when you access / browse the library?

[Ken Maltby]

Originally Posted by Lucas Malor

@Ken Maltby: does your Aura slow down when you access / browse the library?

I haven't loaded but a tiny portion of my library. The books section is only
seven pages at this point. (At six books a page, when I reach 6000 ebooks
it will be a 1000 page menu)

I have created a calibre library: "Calibre Kobo Library" and am at the very
beginning of the process to process the ~8000 ebooks in ~50 categories
(that will become some of the shelves)

I am currently working on 641 books that came from part of one of those
categories in my "Calibre Kobo Library". So I have barely started.

Maybe there will be some options that offer fewer reported problems, by
the time I have several thousand ebooks ready.

Luck;
Ken