Byte, enero del 86

Lo de siempre: seguimos con nuestro proyecto de leernos la revista Byte… con cuarenta años de retraso, y esta vez con un añadido final extra. El tema del mes… ¡la robótica! (Tema que vamos a ignorar bastante completamente, porque no me pone nada. Pero las portadas de Byte son un clásico, o sea que aquí va la del mes:

Portada de la revista Byte de enero de 1986. El tema de portada es la robótica. La ilustración es un huevo, que rompe desde dentro un brazo robótico, como si fuera un polluelo al nacer

Comencemos, pues, por la editorial:

A Threat to Future Software Last October Digital Research Inc. yielded to pressure from Apple and agreed to change its GEM software to decrease its resemblance to Apple Macintosh software. (GEM is an operating environment for several MS-DOS- and PC-DOS-based computers that allows a user to interact with a computer via windows and icons rather than the usual text-only commands.) Let's ignore, for the moment, the uncertain worth of a "visual copyright" (the legal term for Apple's copyrighting of the overall "look" of Macintosh software). Let's also ignore the ethics of Apple's actions. The point to focus on, instead, is that Apple's actions are to no one's benefit: Both the microcomputer industry and Apple itself will suffer from their effects. Apple's actions will slow the growth of the microcomputer industry, which will hurt Apple by shrinking the potential microcomputer audience. Already, several small companies are worried that some project they're working on (and, often, they with it) will be cut down because it is "too Mac-like." In addition, the success of Apple's tactics may encourage other companies to try similar actions, thus increasing the paralysis and anxiety in the industry. These actions will stifle the incremental evolution that is at the root of any significant growth in our industry. By "incremental evolution" I mean the process of gradual improvement of a product type that eventually leads to a more robust, useful product. For example, Ashtonlate's Framework did not spring full-blown from the heads of the programming team at Forefront. It had its roots in Dan Bricklin's and Bob Franston's VisiCalc spreadsheet, Sorcim's Supercalc (which added functions and sold to a market not supported by VisiCalc), Mitch Kapor's VisiPlot (which gave the distinctive highlighted menu bar now used in so many programs), the software integration of Lotus 1-2-3, and the icons, windows, and pulldown menus of— well, you get the point. If companies are afraid to go to market with what they think are incremental— but distinct— improvements on a basic design, we will become a stagnant industry bounded by the usual and comfortable. According to Irving Rappaport. Apple's associate general counsel, Apple's intent is to prevent other companies from creating products that are easy to use because of their similarity to the Macintosh. "If people look at it and say, 'Gee. that's like the Mac— I can operate that,' when that's the result you get, it's over the line" of infringement of Apple's copyrights. The effect of this intent is to fragment the industry in the face of what was becoming a de facto standard for human-computer interaction. This lack of standardization will cause many people to stay uninterested in computers because they will have to relearn basic skills with each brand of computer they encounter. (Imagine how many people would drive cars if car manufacturers used different controls for every function in the car.) Apple might argue that, by claiming a larger slice of a smaller pie, it will still come out ahead. We believe that it will be hurt directly by its actions and will end up with a smaller piece of a pie that is itself smaller. Apple will, in effect, build a wall around its ghetto of Macintosh products, thus limiting its own growth and encouraging people to "live" elsewhere. Texas Instruments' TI-99/4A provides a good example. TI announced that it intended to directly profit from all software written for its machine by forcing third-party software developers to publish their products through TI. When a brave few brought out 99/4 cartridges on their own. TI added a proprietary chip to their cartridges that the computer required before it would run the enclosed software. Needless to say, the few developers working on 99/4 software wisely turned to support other computers. The same may happen to Apple. IBM already sells over half the business computers bought today, and IBM PC-compatibles account for a fairly large slice of what's left. If Apple has been slowing the erosion of its market share to IBM with the Macintosh line (and I think it has), its current moves will alienate software and hardware developers, who will begin to lavish their creativity upon the more congenial IBM PC-compatible marketplace. And where innovation goes, the market will follow. Consider: IBM made its software and hardware architectures open. It allowed the development of innumerable hardware clones, many far more similar to IBM products than GEM is to the Macintosh desktop; consequently, the IBM PC-compatible market far outdistanced its combined competitors in less than two years. On the other hand, Apple is actively discouraging not only copying but also borrowing from its software design. It claims the sole right to benefit from a set of ideas that Apple itself has borrowed and improved on (the most direct borrowing was from work done at Xerox PARC). Given these two opposing directions, what do you think will happen? A Call to Action We at BYTE call on Apple to recognize the long-term implications of its actions and limit itself to prosecuting cases where the alleged theft is not of "looks" but of actual program code. Barring that, we call on Apple to license its allegedly copyrightable interface to markets that do not directly compete with its current or planned product line— if the licensing fees are reasonable, everyone will profit. If neither of these things happen, we call on the judicial system to hand down rulings that reflect a strict interpretation of the visual copyright laws— that is. that a product is at fault only if it shows no distinguishing characteristics in appearance or operation from the alleged original; this would protect products that show incremental evolution. We also call on the industry to do two things. The first is to stand up to Apple and see the case decided on its legal merits. The second is to develop an alternative graphic interface and allow its wide adoption throughout the non-Apple computer community; in this way. the rest of us can get on with the business of making computers— in general— good enough that everyone will want to use them. [Editor's note: Apple maintains that the agreement covers "only three specific products," but one of them is GEM Desktop, which defines the overall GEM environment. Also, according to Kathleen Dixon of Apple, the agreement includes any custom work DRI has done, including the modified GEM software that Atari uses in its 520ST computer] ■ —Gregg Williams, Senior Technical Editor

¿Creíais que Apple se quejaba solo de que Microsoft la copia? (Todo sea dicho: a lo largo de la historia Microsoft ha copiado cosas de Apple… y hasta hay casos en los que Apple ha copiado de Microsoft. Y donde dice Microsoft, puede decirse Google/Android.) Pues antes de quejarse de Microsoft y Windows, se quejaron de GEM, la capa gráfica de Digital Research para sistemas PC/MS-DOS (y no solo estos: volvemos sobre el tema más abajo). Respetando la propiedad intelectual de Apple (más que el editor de Byte, después de leerle), comparto con él que con estas cosas, entonces y ahora, el consumidor sale perdiendo bastante.

Seguimos con los «microbytes» la sección de noticias breves. En esta ocasión, por un lado, evolucionamos con algo que ya habíamos visto por aquí… a las pantallas planas LCD les llega el color:

Epson, Toshiba Announce Color LCDs Toshiba has developed an active-matrix, eight-color, 640- by 480-pixel, 10-inch-diagonal liquid-crystal display (LCD) that nearly matches the brightness of a standard color TV. No pricing or availability information was given. Epson announced a backlit high-contrast, 5.13-inch-diagonal color LCD with a resolution of 480 by 440 pixels (one-third of which are red, green, or blue). Epson says the display's contrast ratio is more than 10 times that of a standard reflective LCD and has a viewing angle greater than 60 degrees. Epson also unveiled a high-contrast, 9-inch-diagonal monochrome LCD with a resolution of 640 by 400 pixels. Samples of both displays will be available during the first half of 1986; prices should be approximately twice as much as standard reflective LCDs. Epson also announced two 10-inch-diagonal monochrome displays using ferroelectric smectic-C crystals. The 640- by 400-pixel and 640- by 200-pixel displays are said to have high contrast ratios, low power consumption, and moderate cost; samples may be available late this year.

Y por el otro (literalmente, hay que girar la página para llegar a ello), desmontamos un poco el mito de que Kodak murió por no innovar en fotografía digital:

Kodak Proposes Tiny Magnetic Disk for Photographs Eastman Kodak, Rochester. NY, has lined up more than 30 companies— including Sony, Hitachi, and Fuji— to support its 47-mm (1.85-inch) floppy disk for storage of electronic still images. The 800K-byte disk can store up to 50 images of 240-line NTSC video. Eventually, the disk is intended for use in cameras; for now, Kodak is working on a 35-mm film-to-disk transfer station for use in developing labs and a still-video player/recorder for the disks.

…y es que pocas compañías investigaron e invirtieron en el campo de la fotografía digital como Kodak, que acumuló una inmensa bolsa de patentes sobre el tema. Lo que mató a Kodak (bastantes años después de 1986) fue, sobre todo, el miedo a canibalizar su mercado «químico».

Nos vamos, ahora, a la publicidad:

Anuncio del modem Hayes Smartmodem 2400

Sí, amigas, 1986 es el año de volar a 2400 baudios, no a los «viejos». Casi dos kilobits y medio, sí. ¿Recordáis la tortura que es tener cobertura «solo» 4G y descargar cosas a pocos megabits? (Pero no os emocionéis: no todas las líneas telefónicas de la época soportaban esa barbaridad de velocidad.)

Y seguimos mirando anuncios, con un momento histórico: ¡el primer anuncio que vemos de Windows!

Gran texto, Introducing Powe Windows. Vemos una pantalla de ordenador con quizás 8 colores y cuatro ventanas, que no se solapan, sino que se muestran una al lado de la otra. También vemos un ratón y un disquet de 5 ¼ con la etiqueta Microsoft Windows.

No os pongo el publireportaje entero (8 páginas tenía en total, que Microsoft ya tenía unos dineros en la época), pero sí os dejo aquí esta maravilla de gráficos:

Doble página con una gran imagen de una captura de pantalla con hasta cinco ventanas mostradas en pantalla, de nuevo sin solaparse. Vemos la aplicación de relog, una ventana con un primitivo explorador de archivos, un "filing assistant" y una gráfica de barras en riguroso blanco y negro.

¿Reconocéis vuestro Windows «de toda la vida»? Yo tampoco.

Hablábamos antes de GEM… y lo recuperamos aquí, porque en este número se analizaba el Atari ST, la tercera de las máquinas con procesador Motorola 68000, después del Macintosh y el Amiga (recordemos siempre: Amiga mejor que ST mejor que Mac). Y el sistema operativo del ST era, efectivamente, el GEM de Digital Research (bueno, GEM era, como con los PCs, la capa gráfica sobre TOS, el verdadero sistema operativo).

The Atari 520ST The 68000 unbounded Editor's note: The following is a BYTE product description. It is not a review— for several reasons. Some of the equipment we received, such as the hard-disk drive, were prototypes, and at the time of this writing, software is scarce. Atari has not yet completed its BASIC interpreter, and the operating system. TOS, remains unfinished. Nonetheless, we are as intensely interested as our readership in new technology, and we feel we have learned enough to share some of the results of our investigations. We began our work on this description as soon as we were able to get a system from Atari. A full review will follow in a subsequent issue. For many years the public has equated the Atari name with arcade games and joysticks. In truth, the Atari 400/800/XL computer line is technically at least comparable if not better than other 8-bit machines, so it should not be a surprise that the company's latest venture, the 520ST (see photo 1), is a competitive 68000 system. Indeed, we are most impressed with the clarity of the graphics, with the speed of the disk I/O (input/output), and with the 520ST's value. The system is not without its problems. The desktop is less effective than the Macintosh's, the keyboard has an awkward feel, and the current operating system makes it impossible to switch between high-resolution monochrome and low- or mediumresolution color without installing the other monitor and rebooting. Nonetheless, we are left with a very favorable impression; several software-development languages are already available, including FORTH, Modula-2, and C. With them, you can tap the power of the 68000 at a most reasonable price. System Description The Atari 520ST is a keyboard computer. Like the Commodore 64 and the Atari 400/800, the 520ST keyboard unit contains the microprocessor, the memory, the video and sound circuitry, and so on. The power supply disk drives, and monitor are external devices. The 520ST has a variety of ports, but there are no internal expansion slots. The In Brief box on page 90 summarizes the features of the Atari 520ST. For $799, you get the CPU, a 12-inch diagonal monochrome monitor, and one external single-sided double-density floppy-disk drive. For $999, you get the same system with a 12-inch RGB analog monitor in place of the monochrome monitor (see photo I). Both systems provide 51 2 K bytes of RAM (random-access read/ write memory), a Motorola 68000 microprocessor, MIDI ports with a transfer rate of 31,2 50 bps (bits per second), a DMA (direct memory access) port with a transfer rate of 10 megabits per second for a hard disk or CD-ROM (compact-disk read-only memory), and much, much more. To be sure, owners will make some sacrifices. The unit does not have an RF (radio frequency) modulator for television output, every peripheral has a separate power supply (wire haters beware), and the operating system currently rests in RAM, stealing over 200K bytes from your workspace. We have summarized other problems below, but almost all are insignificant when you consider what you do get for the money. And rest assured, the system works. Our first system, like most of the first production units, had to have several chips reseated. It now functions properly, and we have not heard of any similar quality-control problems on the latest 520STs. The Hardware Design The heart of the 520ST is the MC68000, with its 1 6-bit data bus and 24-bit address bus, running at 8 MHz (see figure 1). The rest of the system was designed to stay out of the 68000's way. (See the 520ST motherboard in photo 2.) The Atari design team began work on the 520ST in May 1984. From the start, they had several specific goals in mind. The first was to choose a fast microprocessor and do everything to let it run effectively at full speed. To the Atari team, that meant maximizing bus bandwidth and relegating as...

Y… ¿vamos a comparar GEM con Windows, tal y como lo presentaba la mismísima Microsoft en su campaña publicitaria?

Dos fotos de pantallas con GEM a media resolución en una y a alta en la otra. La presentación es muchísimo más sofisticada que la de Windows que hemos visto antes, con ventanas que se solapan y los menús del sistema.

(Eso sí: reconoceremos que el parecido con el sistema operativo de los Macintosh es más que notable. Es innegable.)

Seguimos con nuestra sección «esto no lo ponemos en una revista hoy, que nos lapidan» con un programa en BASIC para dibujar superficies 3D:

EASY 3-D GRAPHICS BY Henning Mittelbach A BASIC program for plotting 3-D surfaces AFTER READING "Budget 3-D Graphics" by Tom Clune (March 1985 BYTE, page 240), I decided to develop a low-cost program for three-dimensional graphics on small computers. The program is based upon the formulas for an axonometric projection in relation to the origin, as shown: XB = X*COS(PHI) - Y*COS(PSI) YB = X*SIN(PHI) - Y*SIN(PSI) + Z Depending on the graphic window of the computer used, you may change these formulas to XB = XO + X*COS(PHI) -Y*COS(PSI) YB = YO - X*SIN(PHI) - Y*SIN(PSI) - Z where XO and YO will represent the origin of the axes, as shown in figure 1. (I developed the program on an Apple II, with XO = 110 and YO = 180.) Also in figure 1, (XB.YB) is the point to be plotted, and PHI and PSI are the angles referring to the horizon. The function Z = F(X,Y), in line 200 of the program, needs a scaling factor F (line 210) that the user has to introduce in the program. The Program The program starts at lines 100 to 180 where you set the parameters X0, Y0, ...

Ojo, que el programa tenía una cierta complejidad y hasta ocultaba las superficies ocultas:

Gráficas de las funciones seno por coseno, exponencial del seno de x por y) y equis por y.

(Si esto no os fuera suficiente, os podéis ir a la página 397 para ver cómo implementar el algoritmo de Euclides para calcular el máximo común divisor.)

He dicho que me iba a saltar la robótica, pero sí me quedo con uno de los artículos de la sección:

MACHINE VISION by Phil Dunbar An examination of what's new in vision hardware THE POTENTIAL APPLICATIONS of machine vision are many and obvious. Everything from quality assurance to robotic navigation could benefit from the availability of reliable vision systems for computers. Perhaps less obvious, though, is the variety of problems that hamper development of the technology. These problems appear on all levels of machine vision— hardware, low-level analysis, and high-level AI (artificial intelligence) manipulation of low-level data. This article will discuss problems that plague the development of vision-system hardware and indicate some of the technology that has emerged to address these problems. You might think that the most difficult hardware problem in vision systems is digitizing the high-frequency analog stream of camera data. In fact, that is not so. Currently, machine vision algorithms use gray-scale (i.e., monochrome intensity) video information almost exclusively. Such information can be adequately extracted from an analog signal by a 6-bit or 8-bit A/D (analog to digital) converter. Real-time conversion requires approximately a 10-MHz conversion rate to digitize a 512- by 512-pixel image. These rates can be achieved with flash converters, pioneered by the TRW company when it introduced the TDC 1007 in 1977. Flash converters employ (2")-l comparators to perform A/-bit conversions. That is, an 8-bit flash comparator requires 25 5 comparators to operate. Since all possible digitized values can be compared to the signal at once, the throughput is much greater than with successive approximation methods. Of course, the complexity of the converter rises exponentially with linear increases in resolution. Notable among the commercially available flash converters is TRW's 8-bit monolithic chip flash converter (TDC 1048) that can operate at speeds necessary for real-time machine vision applications and costs about $140 per unit. The real problems with vision hardware revolve around the cameras. The problems fall into two basic categories: video signal standards and limitations of particular camera hardware technologies. Television Standards Much of robotics suffers from a lack of standards. Machine vision, on the other hand, suffers from the existence of video signal standards that are not appropriate for our needs. Those standards were created by and for the television industry. Since the entertainment industry is still a far more lucrative market for camera manufacturers than machine vision, few image sensors and cameras deviate from television standards. The monochrome video signal standard used in the United States, Japan/ and most of the Western Hemisphere is RS-170, a subset of the NTSC (National Television Systems Committee) standard. Europe uses the international CCIR (Consultative Committee, International Radio) standard, which is similar to, but not compatible with, RS-170. Since both standards present essentially the same problems to machine vision applications, I will limit my remarks to the RS-170 standard. The RS-170 standard defines the composite video and synchronizing signal that your television uses (see figure 1). The image is transmitted one line at a time from top to bottom of...

Y después de la visión venía una pieza dedicada a los sensores táctiles, otra sobre navegación autónoma y una sobre IA en visión por ordenador. De nuevo, uno no sabe si estamos en el 86 o en el 26 (y no se siente con ánimos de explicar a los autores que a la cosa aún le quedaban unas pocas décadas).

Y echamos una última mirada a la publi, y es que creo que no habíamos reflejado por aquí la maravillosa campaña «Charlot» de IBM:

Anuncio a doble página. A la izquierda leemos que el PC ha llvado el rendimiento a una nueva altura. Al la derecha vemos a Charlot sentado sobre una pila kilomética de documentos de todo tipo, trabajando con un PC de IBM.

Que no fue un único anuncio, os lo aseguro. Años, duró la campaña, siempre visualmente maravillosa. Os dejo aquí un recopilatorio de anuncios televisivos.

Y nos vamos a ir con otro momento histórico:

The Acorn RISC Machine A commercial RISC processor by Dick Pountain Acorn Computers Ltd. is one of the U.K.'s most successful computer companies, but like many others, it had its share of financial problems during the depressed year of 1985. Set up in 1 979 by two Sinclair alumni, Chris Curry and Hermann Hauser, the Cambridge-based firm (4a Market Hill, Cambridge CB2 3NJ. England) started out manufacturing a set of modular single-board controllers based on the MOS Technology 6502 processor. These small boards stacked together to make up complete industrial-control systems. The following year the Acorn people launched the Atom personal computer, a packaged but expandable machine that arose out of their experience with 6502 systems. For a while, at around £200, the Atom was the cheapest hobby computer available here, and it attracted a strong following, particularly among those who are as handy with the soldering iron as with the assembler. Hopped-up Atoms can still be found to this day. Acorn's next product, initially called the Proton, was designed to meet a very advanced—for the time— specification published by the BBC (British Broadcasting Company), which was requesting bids to supply a personal computer around which an educational television series would be produced. Acorn won the contract, after a strong and often acrimonious contest in which Sinclair Research, whose 48K-byte color Spectrum was already on the market, lost out. After a frustratingly long delay due to quality-control problems with the ULAs (uncommitted logic arrays), the BBC computer was launched and proceeded to corner the market in schools and universities. Acorn became a very wealthy company, with a turnover reputed to be £100,000,000 per annum at its high point. The BBC Micro (alias the Beeb) is still quite a deluxe machine, with better highresolution color graphics than any of its competitors, and quite a bit faster, thanks to its 2-megahertz 6502. Another plus is the provision of a 10-MHz bus, called the Ttibe, to which second processors can be attached. Acorn charges a lot of money for this sophistication though, and the Beeb has kept its £400 price long after competitors have slashed theirs to below the £200 mark. Acorn had from the start paid more attention to software than most manufacturers, recruiting the brightest Cambridge University computer science graduates for its software division. As a result, the Beeb acquired a range of languages unrivaled by any machine but the Apple II, including an advanced structured BASIC, LISP, Logo, FORTH, Pascal, BCPL (Basic Combined Programming Language), and more. But despite all these positive points, the Beeb has a major drawback, a shortage of memory. The ambitious specification, combined with the limited addressing capabilities of the 6502, left it with a maximum of 32K bytes of workspace (only this year upgraded to 64K bytes), and in the higher-resolution graphics modes this can be reduced to a mere 8K bytes. That doesn't get you very far in LISP or Logo. So at the height of its prosperity Acorn set a team to design, in secret, its own processor to replace the 6502. This may seem like an ambitious, even rash, undertaking, but the people on the Acorn team were so wedded to the simplicity and speed of the 6502 architecture that they found it hard to countenance any of the commercially available 16-bit replacements. The BBC operating system is heavily interrupt-driven, and the sluggish interrupt latency of 16-bit chips, such as the Intel 8086 and Motorola 68000, would have meant introducing DMA (direct memory access) hardware and all sorts of other undesirable complications. Acorn did, in fact, adopt the National Semiconductor 32016 as a second processor for the Beeb, but only after first offering a 3-MHz 6502. And so they conceived the idea for the...

Acorn RISC Machine… A, R, M. La arquitectura del chip de tu móvil. O de tu Mac, si tienes uno. Y ahí estáis, viendo, en riguroso directo, su nacimiento. Casi nada.

Y hasta aquí la Byte del mes. Si queréis hacer los deberes para el mes que viene, como siempre, aquí tenéis los archivos de la revista Byte en archive.org.

Y esto habría sido todo… pero el otro día me enteré de la muerte de Stewart Cheifet (hasta el New York Times le dedicó un obituario). ¿Que quién es Stewart Cheifet? No me digáis que no habéis visto nunca su Computer Chronicles. Si Byte es, al menos para mí, uno de los recursos imprescindibles en formato prensa escrita para revisar la historia de la informática, Computer Chronicles es lo mismo, pero en formato vídeo. Los archivos del programa de la PBS, la tele pública de Estados Unidos (lamentablemente en peligro de muerte, gracias a la administración Trump y su alergia a la información de calidad), son un documento esencial si te interesa el periodo de 1983 a 2000. Y como homenaje, y como estas entradas sobre Byte <ironía>no son lo suficientemente largas</ironía>, he pensado que completarlas con el visionado de los programas correspondientes sería, cuando menos, un ejercicio curioso1. Y os dejo aquí los programas de enero del 86…

El 7 de enero el programa arrancaba con… ¡inteligencia artificial!

(¿No os ha encantado el anuncio del patrocinio de Byte? 😅)

No podemos dejar de comentar el copresentador del programa con Cheifet: nada más y nada menos que el malogrado Gary Kildall, creador de CP/M… y de GEM. Hay múltiples universos paralelos al nuestro en que amamos y odiamos a Kildall, CP/M y GEM y no recordamos quién era Bill Gates ni sabemos nada de un sistema operativo llamado Windows.

El Jerrold Kaplan que sale en la primera entrevista, por cierto, trabajaba por aquel entonces con Mitch Kapor, fundó en 1987 Go, dedicada a lo que luego se llamarían PDAs y luego fundaría el primer sitio web de subastas (cinco meses antes de eBay). Not bad. Y también podemos destacar la presencia del filósofo Hubert Dreyfus dudando fuertemente de la expertez de los sistemas expertos de la época :-).

Maravilloso también que los expertos apuntaban que 1986 podría ser el año del reconocimiento del habla 😅.

Después, el día 14, otro tema del que no se habla nada en la actualidad: seguridad informática.

…aunque en aquel momento esto se refería al uso de ordenadores para perseguir delitos, peleándose con catálogos de huellas digitales o usando sistemas de información geográficos, por ejemplo, pero también digitalizando procesos como en cualquier otra organización.

Os recomendaría, eso sí, saltar al minuto 27:30 del vídeo, en el que Cheifet habla de los gráficos de la peli El Secreto de la Pirámide… creador por un «nuevo ordenador gráfico, creado por Industrial Light & Magic, una división de LucasFilm. El ordenador se llama… Pixar».

Y no sigo porque, según esta lista de episodios en la Wikipedia2, el siguiente no se emitiría hasta febrero.

Apa, el mes que viene más (ya decidiremos si solo con Byte o con el añadido de Computer Chronicles).

  1. Un ejercicio curioso que, inevitablemente, no se me ha ocurrido solo a mí: veo que alguien ha montado un computerchronicles.blog y que ya lleva nada menos que los primeros 133 programas revisitados. ↩︎

Deja un comentario